Full Archon Guide - Build AI Coding Harnesses That Actually Ship (LIVE)

All right, we are live. Welcome everybody to the Archon live stream. We're going to be diving into everything with Archon today.

So, I posted a YouTube video just this week introducing the new version of Archon. Uh, but now we get to actually go deep into using it. And there are a couple of things I have up my sleeve today that are kind of interesting. So, I'm going to be building a what's called a dark factory this month. If it sounds spooky or interesting to you, we'll talk about that more uh in a little bit. Um that'll be kind of like the tail end of our live stream today.

So, yeah, but I want to start by actually introducing Archon to you guys. Uh just to give a whole, you know, overview of the tool. I know that I talked about it on Wednesday in the YouTube video, but I I want to be pretty comprehensive in the live stream today. And then we'll build some workflows together. I'll show you what it looks like to use Archon, how I use it on the daily basis, and then we'll get into the the dark factory stuff. I'm pretty excited.

So, uh, yeah, I'm sitting down right now. I'm like actually like always standing in my live streams and my YouTube videos, but uh I had a couple leg days like in a row. Um, and my legs are like toasted right now, so it's really It actually kind of like hurts to stand right now. It's pretty crazy. Um, so yeah, I'm actually sitting down for a live stream. I've like never done that before. Um, but yeah.

Anyway, so I've got my left monitor here with all of the comments from you guys that I'm watching. And then I've got my right monitor where I have my streaming software. So when you see me look around, that's what I'm doing. And then obviously I'm sharing my screen on the main monitor right here. So, I can actually switch to share my screen here and then I can pop up chats like this right here. Hey, Cole, looking forward to this one. I appreciate very much. Yeah, so I I was checking out the chat as I was getting things started here. It's exciting to see uh everyone in here already. Congrats on the launch. Love seeing the structure for AI workflows. Still wrapping my head around its potential, you know.

I appreciate Thomas and I still am as well because it's actually kind of crazy. Um, maybe we'll talk about this a little bit today, but we can use Archon for a lot more than just AI coding. Like you can use it for any kind of agentic workflow, for deep research or any kind of like content creation. Um, I'm not going to build anything around that now, but that's going to be more coming soon in videos and future live streams. So, yeah, the the possibilities are pretty limitless here. And I think as I really introduce Archon to you guys right now, you'll see what I'm talking about. Cool. All good sitting down. You still look great, Cole. Appreciate it, Eric. Yeah, it feels weird to me, but um yeah, I definitely need it right now. It's exactly right. Never skip leg day. Yeah.

So, so I I started doing so I was doing a lot of CrossFit last year and um I took a break. I just like started going to like a regular commercial gym and I'm now I'm back to doing CrossFit again. And uh the workouts were pretty intense last couple days for the legs. Um so yesterday was a lot of uh Bulgarian split squats, jumping lunges, uh wall balls, and um and biking. So yeah, I'm pretty cooked, but it's good.

Uh, all right, cool. So, I'm going to go ahead and pop up my uh Obsidian vault here. So, I've got the diagram to really introduce why I built Archon in the first place. And I I also want to be clear here. I did not build Archon by myself. There are a couple of people that helped me a lot with Archon. Uh Raasmus and Thomas are the two that have really helped me the most with the project. Uh Thomas is here. He's a DIY smart code. Uh so he's here. I see him in the chat. Shout out to Thomas. Not sure why there's an echo for the stream, Thomas, but yeah, thank you for all the work that you've done on Archon. Rasmus, I don't know if he's here. Uh but yeah, big shout out to him as well. Um yeah, I don't think he is for today, but that's all good. So yeah, I' I've had I've had a lot of help building Archon and of course I've had a lot of help from Claude Code building Archon as well. Uh let me tell you, I am using Archon to build Archon constantly. I'll show you guys what that looks like today as well.

But let's talk about the evolution here. Why we care about Archon? Why why did I care about building it in the first place? So allow me to like talk about this for a little bit. U because I really want to get you guys on board with the vision that I have here. And it is a big vision. So it's everyone knows that like prompt engineering was a big thing when generative AI first became a thing back in 2022. So we had like you know the release of GPT 3.5 Turbo. That's when everyone started giving uh a lot of hoots and hollers about generative AI and people are obsessed with prompt engineering. And the idea behind prompt engineering is how can we craft our prompts to the LLM to get the single best output. So at this point people are very much like caring about just uh you know that very next turn with the LLM. How do we get it to do or output what we want?

And then uh that evolved in 2025 to the idea of context engineering. So now we don't care about just single outputs. Now we care about entire sessions especially for AI coding sessions. Context engineering is the idea of uh you know how can we curate the perfect context nothing more than the coding agent needs but exactly what it needs to uh have all the context it needs to plausibly do the task. So this was popularized by a lot of people like Toby the CEO of Shopify of course Andre Karpathy because whenever he posts something of course it goes viral. Uh but context engineering it's a very powerful evolution because it's treating the context for our coding agents as an engineered resource. So we evolve our AI layer as I like to call it just like we evolve our code base. So we version control our rules and our commands and our skills. And whenever there's a problem that comes up with our um system like there's a bug that our coding agent produces instead of just fixing the bug and moving on. The idea behind context engineering is we look into our process like what could we make better with our commands or what could we make better with our rules so that that issue doesn't happen again or at least that's the goal of evolving our context. So it's a very powerful concept and I created a lot of content last year around context engineering.

Now the important thing is these different evolutions it's not like they replace each other. Context engineering does not replace prompt engineering. In fact prompt engineering is a part of context engineering. So it's it's a evolution where it builds on top of itself. And so this year we have evolved to harness engineering. And this is what archon is all about. So harness engineering can mean a lot of different things, but in essence, a harness is a layer on top of the coding agent. It's the tooling and the process that you build on top of the coding agent to make it more reliable and basically taking your whole you know process for building software and you know building the the layer on top of the coding agent where you enforce that process. So, a lot of different strategies out there like BMAD and GitHub spec kit and GSD, you could consider them harnesses because you're you're wrapping the coding agent in this higher level layer of your strategy for engineering. And we have a stripe minions. Uh the Ralph loop is a harness. Enthropic has open sourced quite a few harnesses. I've covered some on my channel. Really every single big company is converging on the idea of harnesses. Instead of focusing on making the model better or making the tool better, like the coding agent better, it's it's all about like how do we create the system that wraps the coding agent, right? And so there's a lot of studies that have been done that have shown like even if you are using the exact same model, you're not improving the underlying model or tool, you can go from a 6.7% pull request acceptance rate to almost 70%. And the only thing is to harness. So building in your strategies around context curation and validation and your approach for planning, right? Like we we'll talk about all that today when we cover building harnesses with Archon, but it's a big deal. Like uh maybe you guys heard of Stripe Minions. This went like super viral last month because Stripe they built their own internal harness. It's kind of like a more powerful version of the Ralph loop that allows them to ship 1,300 AI only generated pull requests every single week. And like Claude code itself is starting to build a lot of of harness into the tool. So not just calling the coding agent um you know calling the LLM with the coding agent but how can we like orchestrate many different cloud code sessions together. So like claude code has support for agent teams and they're doing a ton with sub aents right now. So when they had their source code leak last month, um it it was shown that like 60% of the code is still like wrapping the the underlying model itself like Sonnet and Opus and Haiku, but then 40% of their code is with these more like harness features like agent teams. And so all that to say, it's the direction that all of these companies are heading where they're building harnesses. The coding tools themselves are building this. All of the biggest companies like Stripe and Shopify and AWS, they're all building their own internal harnesses.

But the problem with all these harnesses is they're not open- source and they're not custom to you. What if you want to take your software development life cycle, your process for working with AI coding assistants and what if you want to package it up into your own harness? Well, that is the value proposition of Archon. Archon is the harness builder. So, for the first time ever, we have an open- source platform that makes it easy for you to build your own harness. Up until this point, you either had to build something internally like Stripe did, but obviously that's a massive amount of effort, or you just had to use a harness that's already out there like the Ralph Loop or BMAD or whatever it is. And like, yeah, those tools are very powerful, but they're not custom to you. And you you guys know if you've been following my content that uh I'm always a big proponent of like build it yourself. Take inspiration from what's already out there. And I'll even show you what it looks like to build something like a BMAD or GSD harness in Archon because we can take inspiration from the beautiful minds that are out there but still make something that works how we want to work. And that's the unlock that I have for you guys here with Archon.

And um the other thing that I want to cover here is um you know everyone already has their own skills and commands and rules. So I'm not I'm not expecting you to like start from zero. I know that you already have a process that you built around your AI coding assistant. Like you have skills for validation. You have commands for planning and maybe you have your PRD template whatever that might be. And so it's not like Archon replaces what you already do. It's that Archon allows you to package everything together into a workflow that combines your skills and commands, right? Like the whole idea is before you have a harness, you have your commands and skills and you have to remember the order that you use them. You have to work between different AI coding sessions because you definitely don't want to do your planning in the same session where you do your implementation because your coding agent builds up a lot of bias. But if you don't have a harness to connect those steps together, you have to be the one to orchestrate that. I call it shephering, right? Like you you go through some kind of PRD process with your coding agent. You create a PRD and then you're like, "Okay, good. Let me go over to a new Cloud Code session or a new codec session or whatever and let me go and create a plan for the first phase." And then you get your plan and then you go and you create a new session and you go into implementation and then you have a new session where you do a review on the pull request or on the code whatever it is. And so you're still even though you have your commands and skills and things to automate different parts of your workflow, you still have to walk the different coding agent sessions through each one. And Archon is kind of like the next evolution of that. It allows you to run longer tasks but still keep yourself in the loop.

And so I have an example here of just what an archon workflow can look like. And and by the way, if you if you did watch the video on Wednesday, I am repeating myself a little bit here. So I covered this diagram in the video, but uh just consider this a refresher if you watched it. And I just want to make sure that for those of you tuning into this live stream who didn't watch the video, I can like give you a really solid overview of what Archon is. Um, so yeah, here's just one example of the kind of workflow you can build with Archon. And uh, you'll have to excuse the the images are like not rendering right now for some reason for some of the icons, but my point can still be made here. Uh, but yeah, so anyway, with Archon, you obviously have some kind of trigger for your workflow. So, traditionally without a harness, it would just be like sending a prompt into claude code, right? Like that is your trigger. But with Archon, there's a quite a few different ways that we can use it, and I'll I'll show you guys that today. We have the CLI so we can have our cloud code or codeex whatever um instance like trigger an archon workflow so we can dispatch work to delegate behind the scenes so we can do a lot of work in parallel. I also have a whole web interface and uh I'll definitely show that off today as well. Uh that's the wrong link. This is it. So we have the whole web interface where we can manage our workflows and we can uh kick off workflows uh from the web UI directly. And uh so there's basically like a a coding agent sitting behind the web interface. So we can ask it to do work on our different repositories that we have registered with archon and it'll automatically route the request. it'll pick the right workflow and it will dispatch it and we can do a ton of workflows in parallel if you want as well. So I'll I'll show you all of that today. But we have the web UI and then we even have different adapters so that you can for example talk to Archon directly in Slack. So you can say like hey there's a GitHub issue for um this repo go and you know handle that issue. I want to see a pull request in the end with a fix for that issue. And you can do that through Slack. And we have different conversations that you can manage through different threads in Slack. So for every single platform, we support parallel execution. Archon handles work trees and isolation under the hood. So you don't even have to worry about it. That's one of the other big unlocks with Archon is how it supports working on things in parallel. And so I can go into my clawed code and I can have it use the CLI to fix eight GitHub issues at the exact same time. and they all run in different work trees so they don't step on each other's toes. They don't override each other's changes and we don't have to deal with merge conflicts just to get the pull requests created. It's really powerful.

And so we start with our trigger and then um again this is just an example of a workflow we can create. So we could go like into planning mode. So maybe there's a new feature that we want to build. We want to start by going through the planning process with our coding agent. And so we have a step in our archon workflow where we are prompting our coding agent like here is the feature we want to build. Now help me plan like do research, do codebase analysis, whatever that is. And when in archon we can add in human in the loop. And so we can have essentially a loop here where we give feedback to have the coding agent revise the plan until we approve it and then it moves into the coding stage. And we do this in a brand new coding agent session because your planning session can get pretty bogged down and you can build up a lot of bias over time. And so you want to produce an artifact that you send into the next node for implementation. If you're in the Dynamis community, this is the piv loop. You know what I'm talking about. I've covered it on my channel as well. The plan, implement, validate. And we used to have to do that between different coding agent sessions, but now we can package this all up as a single archon workflow.

And then after the code is done, then we go into the testing step. And this is another one of the very big value propositions of archon. Our workflows do not have to be just prompts to a coding agent. Sometimes there are steps that we want to run deterministically. We want to take the control away from the coding agent to make sure that our process is followed to a T. If you've used any coding agent for a good amount of time, you know that like sometimes you'll tell the coding agent run the tests after you write the code and it won't listen to that. It's so frustrating. Or it'll do some of the testing like unit testing and linting, but it won't do the end-to-end testing. It's so frustrating. And so what we can do in Archon is we can have certain steps of the workflow where we're just running code like we're doing some kind of context pulling or we're running our tests like we're doing right here. We're guaranteeing that that happens after the implementation and then if there's any failures, we'll prompt the coding agent to fix those things. So we have like a little bit of a feedback loop as well built directly into Archon. And then after everything passes and the coding agent figures that it's done, then we have the human approval. And so even though we are building longer running tasks with archon workflows, we're not taking ourselves out of the loop. We can inject oursel wherever we want in an archon workflow. So no matter how you typically work with your coding agents, if you don't trust it that much and you want to validate it every step of the way, you can do that if you want. We have support built in to the web UI, the CLI, every single adapter for human in the loop. And that's one of the most important things because a lot of times when you have these harnesses, especially like the Ralph loop, for example, the Ralph loop went viral a couple of months ago, but to me it felt like vibe coding, right? Because you were giving it a and it basically an entire PRD, like many different phases of work, and you would just have a coding agent rip through everything. And the problem with that is if the coding agent makes a mistake in the first iteration of the Ralph loop, that issue can kind of propagate and like blow up from the rest of the loop because the coding agent is going to through each loop build on top of a code base that's already wrong, right? Like not aligned with what you actually want to create. And so the issues just compound on themselves. And so the idea that I wanted to be uh very confident in for Archon is that it's not just enabling vibe coding that we actually have like deterministic steps that enforce our process. We have human in the loop, right? And then we have the pull request at the end and then we can review that ourself before we merge as well. And so yeah, that's that's sort of I I like to call it the hybrid secret for Archon.

And um this this sometimes rubs people the wrong way, but I actually think it's smart to take as many decisions from away from the coding agent as you possibly can because they're non-deterministic. They don't make the same decisions every time even if you give it the same prompt. And so there's a lot of inherent risk with that. The biggest reason why a lot of developers and companies are hesitant to adopt coding agents is because they're unreliable by nature. And so Archon is the harness builder that allows you to take the process that you consider reliable and build it into the way you work with coding agents. And so we have different nodes like if you have your testing strategy, if you have your strategy for like pulling information from Confluence, like whatever steps you have that like you want to have performed every single time, you build those nodes into Archon. And then the rest of the workflow is still going to be like you adding in your commands and skills and sub agents. Like everything that you already have, you can build in. So you're not losing anything either. That's another thing that I want to be very clear on here is that you don't have to replace how you already work. Archon just allows you to package it up, right? Like that's the goal of a harness builder.

So I'm I'm so excited for this and I'll I'll actually like install it from scratch with you guys today as well. So, uh, right here I got the link to Archon. I'm going to go ahead and drop this in the chat. So, if you guys want to, um, try it out right now, even like install it along with me, I would highly encourage you to. It's very easy to install Archon because we can ask our coding agent to set it up and I have an archon skill that it loads automatically and it'll walk you through the entire process even installing the dependencies for you. So I'll show you that live in a sec. But yeah, it's so easy to get it installed. Um, yeah, so one other thing I want to show really quick. So this is the readme for the new version of Archon. I want to show what a workflow actually looks like. So, this might look a little bit intimidating, but uh don't worry, you can use your coding agent to help you build workflows as well. I'll also show you that. Um, so this whole like YAML structure, it's it's quite beautifully simple, honestly. And uh but like even even so, like when you're trying to package up your entire workflow, it can start to look a little intimidating. So it's really nice to use a coding agent to help you define these workflows. But just an example here. So within the diagram here, I showed the whole like piv loop workflow. And that that workflow can be summed up with this right here. So this is a a kind of a bit of a simplification just so it fits nicely in the readme, but it gives a good idea here. So you start with the node for your planning and then you go into implementation and you can see that we have we support loops in archon as well. So, uh, maybe I can even go into the UI. Let's see if I have it run with the loop. I want to give like a demo of like what it actually looks like. Yeah, here we go. So, like we have we support loops in Archon. So, by the way, this is what the logs look like. When you invoke a workflow and you view the logs in the web UI, you can see like all the tool calls that your coding agent is making behind the scenes. You can see like the the node flow and like where it's currently running. So like we have a loop right here that has like a certain number of iterations. And if I go to like a more fancy one like a GitHub issue fix, we can see like you can get pretty comprehensive with the workflows that you define. This is one of the the breadand butter workflows in archon uh the fix GitHub issue. So you can throw this on any GitHub issue in any repository and it works through this entire process of uh you know classifying the issue. Is this a a new feature we have to plan or is it a bug we have to investigate? And then it'll do some research. It'll fix the issues after investigating and then it will validate and then create a pull request and review it after. So it's like super comprehensive workflow and and yes having this many steps does take a good amount of tokens. But another really powerful thing you can do in archon is specify the model the at the individual node level. So when you're classifying or investigating maybe you only need to use haik coup and so it might only be the case that like a single node like the fixed issue this is the only one where we'd want to use opus or we'd want to use high reasoning GPT codeex for example and so you can make things very token efficient by determining at each at each individual level what model do we want to use and the other really powerful thing about harnesses like I showed in this diagram here is you get insane results building a harness on top of a model. So the harness elevates the model. I've had better results using archon with sonnet than I have using opus by itself in cloud code. And so uh another really cool thing is like you know this um archon kind of comes with good timing here because anthropic has made their rate limits a lot worse for clawed code recently. It's really unfortunate, but it's forced me to like start using Sonnet more for my coding because I hit my rate limits so incredibly fast. But with Archon, I feel like I'm not really losing the code quality, like the output quality because the workflows, they package up such a comprehensive process and like using this whole thing with Sonnet is still cheaper than like asking Cloud Code to fix an issue by itself with Opus because Opus is just so much more expensive. So anyway, I got off track a little bit showing you guys the web UI here, but yeah, so we have the loop like we can, you know, implement a plan task by task in a loop and then we run the validation. And the important thing is this is no AI. It is deterministic. So we're guaranteed that our validation runs. That's powerful. And then we run the reviews afterwards. And then we have some kind of approval process with a human in the loop before we create our final pull request. And because we have archon as a CLI and a skill, when you have it loaded into your cloud code or your codeex, all you have to do is tell it to use archon. And that like literally that's all your request has to be and it'll automatically dispatch like create an archon workflow. So it'll run like let's say for example this idea to PR workflow. It'll create the work tree under the hood so you can keep working on your codebase in parallel and then it goes through the different stages and you can have your coding agent monitor the workflow and or monitor it in the web UI like I showed right here. So like obviously this is a completed run right now. But uh if this was in the middle of executing you could like see the logs come in in real time and see like you know like what stage of the workflow the coding agent is currently on. And as far as as which coding agents we support right now, we have support for cloud code and then codeex is almost done. And then also we are working on uh or we want to add in support for other coding agents as well. So like right now it's claude and codeex but uh we have a priority item right now to add support for pi as well. So pi also has an SDK. Uh we're also interested in AMP and open code. I mean really like archon, remember archon is the layer above the coding agent because it's the harness builder, right? So we don't really care about what coding agent you're using under the hood. We just have to build a little bit of support for like calling the SDK for the coding agent. So like Codeex has the SDK, Claude has the agent SDK. That's what Archon uses under the hood. So when you use Claude with Archon, it is running Claude code. It's just running it programmatically through the SDK. And um so you can you are allowed to use your anthropic and codec subscriptions with Archon. You don't have to pay for API credits. Um and and that's that's a very important thing because a lot of people right now are getting their anthropic subscriptions banned when they use it with Open Claw and Open Code. But the problem with those they're third party harnesses. They're kind of doing a workaround to use the subscription. And that's against the Anthropic terms of service, but Anthropic has made it very clear. A couple of their team members like posted on X saying that like you are allowed to use your Anthropic subscription with the Claude agent SDK as long as it is for personal use. And Archon is for personal use, right? Like you're running it on your computer. You're hosting it yourself and there aren't other people using your subscript subscription through a production deployed agent. So it's a different story. If other people are using your subscription through a cloud agent SDK agent you have like deployed to some production platform then that's against the terms of service. But if it's a an application you're running yourself and is using the claw agent SDK then you are allowed. So I've been using my subscription with Archon for you know months and months now. And the same thing with my second brain as well. And so they've made that very very clear. I don't have the exact tweet up right now, so I can't show you the exact thing, but uh Boris Churnney, he's the creator of Claude Code. He um he clarified like he made it super clear like yes, you can use your subscription for personal use with the agent SDK. So, we're good. And I get that I get asked that question all the time as I'm doing my second brain content on YouTube and I've you know started showcasing Archon and it is a fair question because um yeah like people are really nervous about getting their subscription banned but you are good with Archon and yeah I see some things in the chat here. People are excited for the Pi support. Uh yeah I'm excited for that as well because then it you will have like support for pretty much running any model that you want with Archon.

All right, cool. Um, all right. So, yeah, I'm going to go ahead and uh answer some questions in the chat here, and then we'll get right into installing Archon. So, I have an instance already spun up uh in the cloud. I I installed Archon from scratch on my computer for the video on Wednesday. I don't really want to like reinstall it on my computer because I already have some things going with it. So, I'm going to reinstall it on a Linux VPS that I created just in Digital Ocean. Uh, but the installation process is going to be the same pretty much no matter your operating system. So, whether you're installing it on a VPS or on your Linux, Mac or Windows system, it's pretty much going to be the same. And the readme has good instructions for that. And I'll I'll walk you through it right now as well. But yeah, before we do that, I'm going to uh switch here to my full frame. Uh, let's let's chat a little bit. So, I'll I'll pop up some questions in the chat here and uh spend some time for a Q&A.

All right. Uh Jared said, "He said the magic words, customize BMAD for inspiration in a custom workflow." That's exactly right. Um, so it works the way you need it, just the way I need it. This is fire. What I've been looking for. Yeah, I appreciate it a lot. Um, and that exactly like that's that's the thing is you can just like fork BMAD and you can just run it and change it yourself, but that's a lot more involved than creating an Archon workflow. I'll show you in a little bit what it looks like to customize um a workflow taking inspiration from something like BMAD. I'm going to use GSD as an example because it's a bit simpler and easier to do, but you could do the same thing with BMAD. Um, all right. So, basically, we could call archons stripe minions but open source. Exactly. Yep. And I'd even go further to say that like stripe minions is just a single harness. Archon can allow you to build anything like Archon allows you to build stripe minions but for your company or yourself. Yeah. Um, can a node be simply an execution node? No AI model used? Yep. Yep. So we have support for running uh bash, python and typescript scripts. So anything like any part of the workflow where you're like I want this thing to happen exactly like no AI model to mess it up then you can run it as a node in archon. Yeah. Uh where does this differ from n? You know uh that's a really good question actually. Here I'm going to switch uh back to my scene here. I'm going to share my screen because if we look at the the builder here and u sorry I'm going to the wrong tabs. We're actually adding support soon for a visual builder. So you can like connect the nodes like this. And so it is going to be kind of similar to N8N. In fact, in the readme here, we say think N8N but for software development. And so for those of you who use N8N or have used it in the past, this might actually click really nicely for you is like N8N doesn't really have direct connections with cloud code and they don't have things like workree isolation support, right? Like you don't use N8N for AI coding. You use it to build automations for your business, whether it's like LLM workflows or not. But u Archon is going to allow you to build kind of like n workflows, but specifically for AI coding. So you're stringing together these coding agent sessions and deterministic steps like bash and python scripts. And it's for coding processes. Uh, and so like there's a lot of things we have built in behind the scenes with Archon for how it integrates with our different coding agents and handles work trees and things like that that N8N doesn't have at all. So like there's no way you could build this kind of thing in N8N. You're you're not going to be able to build a coding agent harness in N8N or at least it'd be very difficult and not really what the tool is meant for. Um, so this is a very different use case, but you can think of it like N8N for AI coding. So yeah, I appreciate you asking that. All right. Um, how can we know that the plan step is complete? Do we have human an approval gate for that? Uh, yes. So, you can have the coding agent decide itself that the plan is complete or you can make it so that you have to approve. And so, like going back to the diagram here, this step right here is what you're talking about like planning. And so what it does is it'll output some artifact and you can have I know that I only show the human approval gate here, but you could like add that as a node right here as well. So like you get to look at the artifact and it'll like give you a summary of it or it'll give you like the full file. So you can like go through the markdown if you want and then you can give feedback. So you can say like hey you didn't do enough validation, right? So or like you didn't add enough u planning around validation. So like go back edit the plan and then let me review again. So you can add that for any step in Archon or you can have it just decide itself. Like if you want it to be a little bit more hands-off, like you want to do something like the Ralph loop, which by the way, we have a Ralph loop workflow built into Archon for you to use out of the box. Um then you don't have to have the human in the loop. Like you can have it kind of like iterate by itself and kind of critique itself and then decide when it's ready to move on, which obviously that's a bit more or a lot more non-deterministic, but you can do whatever you'd like.

And um speaking of that, actually like workflows in Archon, we have a lot of Archon workflows that come shipped with the platform. So when you install Archon, all of these workflows here are ready for you to use immediately. So you can build your own. If you want to create your own harness, build your own workflows, you can. But if you just want to poke around with Archon initially, I have all of these ready for you guys to use. So we have like a human in the loop interactive PRD to like walk you through creating PRDs. Uh we have the whole like plan to pull request. We have the Ralph loop one if you want to do the Ralph loop in Archon. It's actually like a beautifully simple workflow. Most of it is just prompting here, but it'll walk you through like creating the PRD. It'll validate it and then it'll go through the whole Ralph loop and like manage the state through uh like JSON and markdown files just like the Ralph loop does, the original Ralph loop. Uh what else do we have? We have the uh GitHub issue fix. This is the one that I use the most out of all of the workflows in Archon because I'll I'll throw this on the issues that we have in Archon. Like I'll I'll say like, "All right, Claude Code, uh, we have a few people that have just opened up issues for Archon. We got, you know, like 10,82, 80, 76, 72. I want you to spin up four GitHub issue fix archon workflows in parallel and then monitor them and let me know when they're done and I can review the poll request." Or I could even have it like run the validate PR poll request after as well. Um, so yeah, like I said, I'm always using Archon to build Archon. All right, visualization is fantastic. I appreciate it. I assume you're talking about the the web UI here, but yeah, there's a lot of work that we've been putting into the visualization here. Um, can I use this with open code? Uh, yeah, so not yet, but we want to add support for open code. So pretty much Archon is going to be able to be integrated with any coding agent that supports an SDK. So like like I said with Cloud, we have the agent SDK. Codeex has their SDK, PI has one, Open Code has one, and AMP has one. Uh Gemini CLI does not. That's another one that I wish I could add, but they don't have an SDK. Um now you can just run the coding agent in headless mode, but uh it's it works better to use the coding agent programmatically. So this entire codebase is Typescript almost entirely TypeScript because that's the language most of the SDKs are written in. So the cloud agent SDK has both a Python and TypeScript version but Codeex is only TypeScript. So that's why I decided to go TypeScript for for this codebase here. Um and in the end like the actual language you pick doesn't matter a ton because coding agents can just rock everything, right? In fact, Typescript is even a bit better for coding agents than Python because um it has the type safety built right in. All right, cool. Um All right. So, yeah, there are so many good questions in the chat here. It's going to be hard for me to get to everything, but I'll answer a couple more here and then I'll get on to the demo where I'll I'll install it from scratch with you guys and show you how I use it on the day-to-day. Um, why have you made this open source? That's a good question. So, open source has always been a really big part of my ethos. So, as I created started creating YouTube content um in 2024, I I really just had a passion for sharing my knowledge and everything that I'm diving into with the whole world. And open source is my way to share it the most I possibly can because otherwise if I close source it I'm going to tell you guys about it but then you won't really care because it's either going to be some paid product that you have to shell out a bunch for or it's meant for enterprises or whatever and then like no one is going to care about it. So, open source is my way to get the most eyes on it, which it also is helpful for me because then I get the most feedback for it. And um obviously like for my channel, it helps because then everyone has a reason to care about it. If it's not something that's open source, I can share what I'm building, but then maybe if you want to like follow along with my journey, that's interesting. But otherwise, it's like nothing you can really try yourself. So, I feel like it just you can be kind of like a brick wall you hit. if I make a YouTube video on Archon, it's like look at this cool thing I built and then you can't even try it. It's like what's the point? U so definitely like with YouTube being my primary like platform for sharing things with the world like what I share has to be open source in my opinion. So yeah, it's a fair question though. All right. Um, yeah, so another person asked about this in NN. So, I talked about that already, but yeah, I appreciate you guys asking about that. Uh, will it include the GitHub CLI, GitHub Copilot CLI in the future? So, if they build an SDK, yes, but I don't think they have an SDK. Um, oh, wait a second. Do they actually look at this? When was this released? They Oh, okay. So there is a SDK. So actually we could integrate GitHub Copilot with Archon as well. That's pretty cool. And yep, they support TypeScript, which is what we would need for Archon. So yeah, this is yet another one. And okay, here's the really cool thing. I I haven't really talked about this yet too much. I mean this gets more technical but um let me let me tell you there is so much effort that I put into architecting the initial codebase for archon before I I ever wrote a single line of code. So I have like a sort of like generic interface implementation for every single adapter and every single coding agent. So when I built the initial version of archon I supported claude and I supported telegram. So like Telegram was the way to talk to it remotely and Claude was the only coding agent but I built it in a way where it wasn't like super coupled with that specific tool like Telegram or Claude. And so when I asked Archon to build support for Slack as well as Telegram or or GitHub we like support you can like talk to Archon directly and GitHub issues as well. It just one-shotted it like I didn't even have to iterate. it built it perfectly and then when I asked it to build support for codeex like following the pattern that I did for claude it oneshotted that as well and so if you wanted to create a poll request adding support for GitHub copilot you could you could literally oneshot it because the documentation is there the coding agent can reference that if you have like claude code open up in the archon repository or codeex and um and then also like it can just copy how we already have it set up for an existing coding agent. So we have Archon set up in a way where the codebase is very easy to evolve because everything is set up as like super easy patterns to understand like the coding agent can understand very easily like here's how I add a new coding agent here's how I add

A new adapter, uh, whatever we want to do to evolve the system. Um, can workflows be told to obey token rates and limits? So, I think that's actually something we have an open issue for, cuz there's not a way to like set a max budget for tokens, but it would be pretty easy to add that like on a per-node level. It's like, I want you to stop if this is taking more than 100,000 tokens, cuz this, that definitely means that like the coding agent is going off the rails here, or whatever. So, we don't have support for that yet. But, um, I think, I mean, I, I don't really want to like go on a hunt here and try to find this exact one, but, uh, there are, there is definitely an issue that we have out for, for that. Let's see if I can find it.

So, most of these issues that we have right now are actually created by, uh, Raasmus. So, like I said, he's one of the guys that's been helping me an insane amount with Archon. So, most of these things are like just us kind of listing out things that we want to, uh, improve in, in Archon. Um, we had to port over. So, we were working on kind of like a private repo that was like just for the Dynamus community at one point. So, we had to port over a lot of the issues. I don't know. We might have changed the names for some of them, but that's definitely something we have on our radar. Um, yeah, not sure if I can find it exactly right now, but yes, that's one of the things is just like all of the different parameters that we have in like Claude Code and Cloud Code. We want to make sure that we can support those. So you can build it directly into the configuration for each node of an Archon workflow.

All right. Yeah. And so, yeah, one thing I, I, uh, want to mention really quick is like when we were first working on Archon, we were doing it like as an internal project in the Dynamus community. So, we called it the, actually, I still need, I still need to archive this repository here, but we have the remote coding agent. It's funny because Archon, the new version of Archon, started as a, a resource that I built for the Dynamus Agentic Coding course because it originally wasn't a harness builder. It was just a platform that allowed you to talk to Cloud Code or Claude Code, uh, in a remote environment like Slack, GitHub, or Telegram, like I was talking about earlier. So, that's how it started and then it evolved into this beautiful thing where now you can like build any AI coding workflow as an Archon harness. Uh, but this, this is like the origin of Archon. So, it's cool for everyone in the Dynamus community. You guys got to like see it evolve to the point where it is now and obviously you got early access to it.

Another thing is, um, within the Dynamus community, I am going to be doing a lot more workshops the next couple of months, like getting like really deep into using Archon, like building custom workflows. I'll show you guys more like how to build your own harnesses. Um, it's, it's always the place to be if you want to like be a part of Archon and get the the inside scoop on how to use the tool the best and like the evolutions that I'm doing to it. U, so, yeah, I just want to call out really quick for Dynamus. I'm going to put a link in the chat here. Uh, I actually have a little live stream special for you guys for Dynamus. So, it's, uh, 10% off the, uh, public price for the community. This discount here is going to go away, uh, by the time our live stream is done here. So, it's a special for literally just this live stream. So, if you like what I'm working on with Archon, if you want to be a part of of this and, um, also all the other course content that I have in Dynamus, definitely check this out. I put the link in the chat just now.

So, another big thing that I did in Dynamus recently is I did a four-hour boot camp on building your own AI second brain. So, I took the entire second brain that I built for myself that literally saves me 20 hours a week and I showed you how to build it from scratch. So, I did a, a live workshop and I'm also turning this into the third course for Dynamus as well. We also have the AI Agent Mastery and Agentic Coding courses and I referenced the Agentic Coding course already because that was the origin of the this new version of Archon. So, yeah, lot, a lot of value packed into Dynamus. I do weekly workshops as well. And so, that's where I'll be doing some more stuff with Archon if you are interested. So, all right.

Uh, with that here, I'm going to go ahead and show you guys how to install Archon from scratch. And, uh, like I said, I'm going to be doing it on a VPS just because I don't want to wipe my installation on my computer again. I, I already reinstalled Archon from scratch, um, four times this week as I was testing things to get ready for the open-source release. And so, I'm going to install it on the VPS where I'm going to build my dark factory. This is the exciting, like, last part of the live stream that I'll be covering with you guys here. Um, yeah, I'm, I'm actually pretty stoked for this. I don't know if you guys know what a dark factory is. Um, it's kind of like a, a term that was coined in the last, well, actually it was a, a term that was coined in the late, uh, 1900s. Uh, that sounds weird. Like 1990s, I think. Like the idea of a dark factory is, um, you have a factory without lights because it's robots running the entire thing. And then recently people have been talking about like dark factory as it relates to code bases and basically a dark factory is a code base that self-evolves. Like AI is the only one writing code ever on the code base. And so an experiment that I want to run, I'm giving you a little bit of a teaser to what I'll talk about at the end of the workshop here. An experiment I want to run is a public dark factory, factory code base where every single evolution of the code base, like every pull request, every release is managed by Archon workflows because we can build our process. Like we can define exactly how we want to manage issues and pull requests and releases. We can define it as Archon workflows. So, we basically use Archon as a dark factory harness, which sounds kind of, it sounds silly, but I'm actually like really excited to try this as a, um, kind of like public experiment here. Like I, I literally wanted to get to the point where like anyone can create an issue and then I'll have Archon like figure out, is this an issue that we should address for this code base? And then it'll handle it automatically all the way to like pull requests and reviewing and merging into the main branch. So, we'll talk about that, but obviously we got to get Archon, uh, spun up for the first time here. So, I'll show you guys the installation. I'll show you what it looks like to run workflows with Archon, and then I want to build a workflow with you guys, and then we'll do some of the dark factory fancy stuff at the end.

All right, so I'm going to go into the readme here. So, I'll put a link to Archon in the chat again. If you guys want to follow along and install with me on your machine or a VPS, if we scroll down to the get started, this is where we're going to work right now. So, literally all you have to have installed before you run Archon is your coding agent and the GitHub CLI. In fact, I list Bun as a prerequisite, but the setup process is going to even install this automatically if you don't have it yet. So, I do list Claude Code as a prereq. You can use Claude Code as well, that's also supported with Archon. It's just not quite as stable right now as Cloud Code. So, I'll definitely be updating the docs as I address that or as we address that. But then obviously the GitHub CLI is an important, uh, dependency as well. You don't need the GitHub CLI for every Archon workflow, but, um, yeah, being able to work with GitHub is a very core part of Archon because most workflows are relying on having the issue or pull requests be like the starting point, right? Like reviewing pull requests or fixing GitHub issues. Most of the time when I'm working with Archon, it's, it's dealing with artifacts in GitHub essentially, right? So, that's why I have that as a prerequisite.

So, once you have those things installed, um, we have the binary. So, we're working on making it so there's like a single install script. So, you can do this if you want. I'm just going to show you that my my approach of like cloning the repository and setting up from there, cuz that's what I'm more comfortable with. This is like a newer thing that we're working on to make it like even easier to set up Archon. So, I'm going to install it this way. So, I'll, I'll copy this command to clone the repository and I'll do that within my VPS here. So, there we go. Clone Archon and then obviously change my directory into Archon. And then check this out. So, I'm going to run in dangerously skip permission mode just because I don't want to have to deal with approving things right now. Open up the folder. Uh, yes, I accept. There we go. All right. Now watch this. All I have to do is say setup Archon. I don't have to say anything else. It's going to walk me through the entire process. So, if you have the prereqs and then you clone the repo, you just go into your Claude Code, Claude Code, and you say setup Archon and that's it. So, first it's going to see like what we already have set up. It's going to realize that we haven't set up anything. And then what it should do is it should load the Archon skill. So, we have this skill that like walks it through how to help us with Claude or with the, the whole setup of Archon with Claude. All right, I guess I have to wait for it to to work here. And by the way, if we see it, um, I just did like control O to see the output here. It's just like understanding the code base. I'm actually surprised I didn't load the Archon skill. Um, you're in the, let's see. What is it saying here? No. All right, hold on. Load the Archon skill and walk through the setup. I've actually never had to do this before, but it, uh, for some reason didn't load the Archon skill. So, I'm just going to tell it to do that, but we're still going to get the same effect here. It'll be super easy to go through everything. All right, there we go. So, now it's reading the setup guide. So, this is like a part of the skill that, uh, tells it how to walk us through everything. So, it's going to check all the prerequisites like making sure we have Git and Bun, things like that, which we already do on this machine. Um, there we go. So, now the first question that it'll ask you in the setup is where, like, what's the repository that you want to use to work on with Archon, right? So, like when we use Archon, we want it to invoke workflows on another repository because we're going to use it to work on something else. So, that's why it says this should be your own project, not the Archon repo. And so, I can either clone a repository from GitHub or I can just give it a local path if I want it to work on a, a project that I already have installed on my machine. So, number two is actually probably what you would end up doing because usually there's already something you're in the middle of working on that you want to use Archon to like build a harness around, right? Um, now since I'm on a VPS, I am just going to clone something from GitHub here. So, I will select that. And then it just says like, "Please provide the repo URL." Um, I need to actually find one here. I'll, I'll just do, let me open this up. I'll pull some like random, um, let's see. Let's just do this one. I'll just copy this rep. I just need kind of like a random one here to register. So, what this is going to do is it's going to register your first repository with Archon. So, whenever you, um, run the Archon CLI on a repo for the first time, it will register. So, like Archon has a database under the hood or behind the scenes that like keeps track of the projects that you're using Archon with and like all the conversations you've had where you've invoked Archon workflows on those code bases. So, you do the automatic registration that way. And then also in the web UI, like I showed earlier, you can register projects this way too. So, you just click the plus icon here and then you can give it the GitHub URL or the local path. So, also in the web UI, you can register your repos. And the cool thing is for all the projects that you have registered in the web UI, the agent automatically understands. So, if I say like, what projects and workflows do I have, the Archon agent that runs in the web UI, either using Cloud Code or Claude Code, it, it has the context injected for all of the Archon workflows it can invoke and the different projects that it can invoke it in. And so, when we say like, hey, I want to fix this GitHub issue for this project, it will know like, okay, let me use the fix GitHub issue workflow and I'll route it to the rag YouTube chat repository. And the, the web UI is one of the things that we can spin up once we go through the setup here.

So, first it asks what platforms do you want to set up? So, the CLI is always included by default. I can also set up, uh, you know, like, uh, Telegram. You just do, you know, like, um, enter to select or unselect. Uh, for this, I'm going to keep it really simple and just do the CLI. Um, yeah, but you, you can set up like all the adapters that we support in Archon right here. So, you specify the ones that you want and then in a little bit it'll go through a process where we can give our API keys, but it's a separate process because we don't want to just send our API keys directly into a coding agent or an, an LLM. So, I'll submit these answers here. So, I just want to set up the CLI so I can run workflows there. So, the CLI and the web UI obviously don't require you to have any additional parameters or API keys. So, they'll just come, they'll work right out the gate. So, then what it does here is it creates Archon as a global command. So, that way we can just run, you know, like Archon, uh, workflow run, whatever. So, we can use the CLI to invoke workflows. But trust me, you're never going to do this yourself. You're just going to have your coding agent run Archon workflow. So, you don't really have to understand the CLI yourself. That's the beauty of having an Archon skill is your coding agent can load the skill and then it knows how to invoke the Archon CLI to kick off workflows for you.

All right. And so, now what happens is it helps you configure your credentials, but it runs this in a separate terminal because we don't want our coding agent to see our API keys. That would be a huge security risk. And so, it, uh, it will automatically spin up a new terminal for you to go through the setup, but depending on your operating system, it might not, there might not be support to like automatically start a new terminal. And so, uh, you might need to just use, do this yourself. And so, for me, because I'm running in a VPS, there's no like automatic terminal spin up. So, I just have to connect into the machine again. So, let me do that. So, I gotta, uh, remember the path to my SSH key. Uh, there we go. Goodness, I can't type today. Gosh. All right. And then, um, I forgot the IP of my machine. Shoot. What is the IP address of my machine here? I'm going to pull this up on another monitor here. Um, all right, pull this up. There we go. Okay. All right. So, SSH in and then I will go to, uh, the path here. Dark Factory Archon. Wait, that's not it. Dark Factory. Uh, oh, yeah, capital A. Okay. All right. So, then all you have to do in a separate terminal is run the Archon setup command and it, it walks you through this here. If it doesn't spin up the terminal automatically, you just do that yourself. And now we just go through the setup process. So, I'll zoom in on this here. So, first, what database do you want to use? SQLite is the easiest to set up. And so, I recommend that. You can use Postgress though if you want to have an external database you connect to. Uh, which, shoot, I accidentally entered that. So, let me go through the setup again. So, SQLite and then here's where it asks what coding agent you want to use. And so, these are the two we support right now. Like I talked about, we want to add more as well. So, space to select and then enter to confirm. And then it asks, how do you want to authenticate with Claude? And so, there's three options here. We can give an oAuth token, like if you want to just run, you know, the Claude, like setup-token command and get an oAuth token. You can give your API key, which I wouldn't recommend because it's going to get expensive. And then, like I said, we are allowed to use our Anthropic subscription with Claude Code with the Claude SDK and Archon. So, I'm just going to use my global oAuth. So, on this instance, I've already authenticated with Claude. If you use Cloud Code a lot and you're installing Archon on your machine, then the authentication is already set, right? Like you don't even have to set up any other environment variable. And then it asks what platforms do I want to connect? As in, what ones do I want to enter in an API key for? I'm not actually going to, I mean, I guess I'll just do GitHub right now just so I can show you one of these. So, you do, um, space to select and then enter to confirm. And then what it'll do is for each of the platforms that you select that you want to install Archon with. This is very similar to the, um, OpenClaw setup. If you guys have installed OpenClaw before, is for each one of the platforms, it'll like give you instructions like, here's how you get your, your GitHub personal access token. And so, I'm going to, uh, off camera obviously go ahead and, uh, get that and copy it. So, let me go to, there we go. So, I'll paste it in because it automatically hides it, which is good. I would want to show that on the live stream, but yeah, it gives you instructions for how to do that. It'll be the same for every single one of your platforms that you configure. And then another thing, obviously for the sake of security, is we, for each of of the platforms that we set up, like GitHub and Slack and Telegram, we want to have a comma-separated list of users that are allowed to invoke Archon because if I have Archon running on a public repository, I don't necessarily want anyone to be able to just say like, you know, at Archon fix this issue, because then it's spending my tokens and it's not me invoking it. So, I'll just say like, this is the only user that's allowed to use it. And then you can change the mention name. So, like in GitHub, you do @ Archon and then you give it the request, like that's how you talk to it. So, I'm good with with that as the default. Um, so I don't need to set up anything separate. And then the other thing is that also for that repository that we've, we registered. So, this repo that I found just as like a, a random one. Where was it? Yeah, this one. So, like I, I gave it this to register as the first repository. Um, we can also copy the Archon skill into that project. And the, the reason I want to do that is then I can open up my coding agent directly in that code base and since the Archon skill is there, it knows how to use the Archon CLI. So, that way I don't have to open my coding agent in the Archon repo in order to work on another repository. Because as long as the Archon skill is there, the Archon CLI is a global CLI. So, we can invoke it from anywhere on our machine. Like I can just open up my terminal right here and just say like Archon, right? And then that automatically works, uh, because it's a globally registered command. But our coding agent only knows how to use it, like use all the different commands and options if we have the skill. So, I just want to copy it over into, uh, the code base. So, all right. And then, uh, the docs directory. We don't have to worry about that right now. And there we go. Our setup is complete. And so, we go through all of this and then we go back here and we just say done. So, like we finished and, uh, now it's going to validate all the credentials and make sure that we're good to go. And then it'll actually test the Archon CLI. So, it'll run a workflow for us to make sure that everything is all configured properly. So, we'll let that run here as well.

All right. So, there we go. Running a quick test here. So, we can see that it is using the Archon CLI. So, thanks to the skill, it knows how to. You never have to worry about running it yourself. You just ask the coding agent to do so. So, it's running the Archon assist workflow and it is doing it on our, uh, repo that we registered, the first repo that we registered. And so, it's just, you know, say hello, like a really basic test to make sure that it worked. And there we go. It's good. So, yep. I'll copy the skill over. I guess it asks us here as well. Maybe that's something to touch up. Uh, but anyway, so every single time that you invoke a workflow from the CLI, it's going to run as a background process. So, Cloud Code or Claude Code, like it has access to, um, the full logs from the workflow. So, it, you can like ask it for a status if it's a longer running workflow or you can, you know, say like, "Hey, summarize what happened in the workflow." Like, it, it's able to basically, you know, communicate with the workflow that it runs because all the logs are right there in the background process. So, there we go. Our setup is complete. So, it tells us what is configured and then it gives us next steps as well. So, take a look at this. We are immediately ready to use Archon in our code base now and we can register it with any other code base we want as well. So, like I could just say like, you know, I'll go into my speech to text tool here and I'll say like, use the GitHub issue fix workflow to fix issue number two on, uh, XYZ repository, right? Like I can send off this request and it will automatically, if, if I, you know, give it obviously the path to the repo here, it'll use the Archon CLI to run the GitHub fix issue workflow on this repo and then that also automatically registers that repository with Archon. So, it knows about it going forward if the repository wasn't registered already. And then the other thing is wherever I have the Archon skill copied, which if you want to copy it yourself in the Archon repo, it's just within .claude skills. So, this, and, and I'll put a link to this in the chat right now as well. This skill, which you can just ask your coding agent to copy it into your new, like, whatever repo you want to start using Archon with. But this skill tells it how to use the CLI. Like, this is literally the only requirement. So, another thing is if you want to use your second brain with Archon, all you have to do is put this skill into your second brain repo and then it'll immediately be able to start using Archon workflow. So, you can basically add Archon as the coding arm for your second brain. So, you can create your own harnesses, your own workflows, and then you can tell your second brain to invoke it on whatever repo you, you want. As long as you give it the path, and it loads the Archon skill, then it knows how to use the CLI. And so, for anything here, if it's like confusing, like, okay, what repo do I open up or like, how do I get the skill or how do I register projects? Like, you literally just ask Archon and it knows how to do everything because the skill walks it through everything.

So, let me actually show you this here. So, in this other terminal here, I'm going to, um, clear and then, so it says I can change my directory into the Claude memory compiler and then I can run Claude to launch Claude here. So, I'll do, uh, the dangerously skip permissions again or I guess I don't here. I'll just have to type it out manually. So, Claude, dangerously skip permissions. And then, um, you can see that since I copied over the skill to this repo during the setup, I can just say load the Archon skill. Not that you have to say this explicitly every time, but I'm just demonstrating that like we have the Archon skill in this repo. So, now it's going to know how to invoke any, uh, workflow. And then within the repo here, okay, we actually have a couple of issues. Cool. Uh, this 6 PM thing is very edgy. I've never actually seen this issue before, by the way. This is a public repo. Um, that's okay. Massive token consumption. Interesting. All right. Well, I'm trying to find one that's actually like, um, seed existing. Okay. Well, let's try this one. This is kind of random, but I'm just going to pick a random issue like number one here. So, like watch this. All I have to do is say, uh, I want you to use Archon to fix issue number one. That's it. It's so simple because it knows the workflows it has access to. It's going to pick the right one, right? Like it's going to pick the Archon fix GitHub issue. It knows how to use the CLI. Boom. That is it. And we have this full process running now. So, Archon fix GitHub issue is running in the background. It's going to check on the progress periodically. And, um, oh, hold on. The workflow failed because you're not logged into the GitHub CLI. Oh, that's a bummer. Okay, hold on. I guess I have to do that. I thought I already did the login here. Um, let's see. Paste authentication token. Hold on. I'm going to do this off camera quick. Um, I thought I already did that part of the setup, so it's kind of weird that it says, but I'm just going to try the login here. All right. So, yeah. See, it says I'm already logged into this account. I think Claude might be tripping right now. Uh, let me try resuming the conversation. Let's see. Uh, I logged in. Also, it says I already was. So, I am confused. There might be something else I forgot to or I messed up in my configuration on this machine. Um, let's see. But it's cool. Like it can just rerun the workflow, right? Like we can just talk to Claude as we normally would and so it can use Archon as a tool just like it would use sub-agents or just like it would use skills. Uh, it says it exited again. Um, the error is actually from the Claude Code off the not logged in is coming from the Claude Code agent that Archon spawns. Um, the check Cla, oh, I think I know what it is. I think it's because of the, uh, it's because of this specific repo. I have the Claude folder, the settings.json. If I just remove this, I might just do the demo on my computer instead of this VPS because I, I think it's just this specific code base that I have some like Claude Code hooks that are running. It's a whole thing. Oh, wait. Oh, wait. I know what's wrong. It's because I have this already set up to use Minimax M2.7 for the dark factory stuff I was going to show you guys. That's my bad. Ah, okay. I have to, I have to demo this on another machine here. By the way, we were using Minimax 2.7 for everything. Um, because I was getting some stuff set up ahead of time for the live stream here. Sorry guys, I'm complicating things more than I, um, than I need to because I, I have some other things prepared for you guys here. So, okay, here's what I'm going to do. I'm going to show you a demonstration by doing it right from Archon. So, um, this will be better anyway because then I can show you guys the web UI. Uh, which, by the way, if you want to start the Archon web UI, all you have to do is go into the Archon code base and say start the backend and frontend of Archon. That's it. Um, start the backend and frontend of Archon. I guess my speech text got cut off, but that's all you have to do. So, I hope that like the pattern is clear here for your setup for getting things up and running for running Archon. It just comes down to like the Archon skill guides it through everything, all workflow execution, managing the application, super, super easy.

So, within my Archon here, let's go ahead and, uh, zoom in a bit. Let's say I just wanted to handle an issue. So, this is going to be kind of meta because I'm using Archon to improve Archon. Um, but like I said, you could do this on any code base where we have the Archon skill. Um, so I'm going to find an issue. There's quite a few issues that have been created in the last couple of days because we're, we have a lot of eyes on Archon right now. We're at 16.2,000 stars, which I'm honored, by the way. Um, okay. Uh, let's see. So, okay, this is a good one. So, chat UI fails silently when the Claude oAuth refresh token is expired. So, this is issue number 176. So, watch this. I'm gonna go in this just like I was trying to do in the VPS. I just have things, uh, configured. We'll talk about the dark factory stuff in a bit because I want to use Minimax. Um, anyway, so I'll go in here and I'll say, uh, use Archon to fix, uh, issue number 176. There we go. All right. Now, now we'll see it in action. So, uh, again, it'll load the Archon skill. Well, I guess first it'll view the GitHub issue and then, yep, load the Archon skill so it knows how to invoke workflows and then it'll kick off the workflow. So, Archon fix GitHub issue and then, uh, we're doing it in a branch. So, it's going to do this in a work tree. So, we have isolation and there we go. So, now the workflow is running in a background process. So, Cloud Code has support for this. I don't know if Claude Code does as well because I haven't used Claude Code in a while, if I'm going to be honest. But if you click into the shell here, like if I, if I press the down arrow and then hit enter, I can see the details and the logs of the workflow as it's running. So, we can monitor it here. And then the cool thing is because it runs as a background process, I can continue. I can run more Archon workflows. I can keep just talking to the agent here. I can also say, give me a status update. So, it can look into the logs for the background process and then tell me what stage of the workflow it's in. So, if I want to check in because this is a longer running workflow, it can say like, oh, it's currently investigating or it's in the middle of classifying the issue. So, it reads like, I know it looks kind of long here, but this is the, the logs that are stored internally on my machine for the background process and then it says, all right, cool. So, work tree is created, it extracted the issue number and it classified the issue. So, it's currently running the web research step and if we go to the web UI here, I can actually see that. So, take a look at this. We have it currently running. So, I can see it in my chat. If I go to the mission control right here, I can also see all of my running workflows at a high level. This one is currently running. It has been for a minute and a half. And if I view the logs, take a look at this. I can see what steps have completed, where I currently am, and I can see the tool calls as they come in, which a lot of times for Archon workflows, you're just going to fire and forget, right? Like you just want to have it handle something in the background, and then you'll come back once there's a pull request for you to review. So, it's not like you're always going to be watching your workflows in the web UI, but especially as you're building your own custom workflows, it can be really, really useful when you're first debugging things to dive into a workflow log and like make sure that things are actually happening as you intend them to. And so, we can see the logs. I know like this one specifically, the, the web searching, it, it's, um, a longer step, but like we can see all these tool calls come in live as, uh, as it's working. Um, and then we can invoke a ton of different workflows in parallel and we can watch them all here. We can click between the logs for them. Uh, it's pretty cool. So, yeah, and so we can monitor it here or in the web UI.

And, uh, you know what? Just for the sake of of getting kind of fancy here, I'm going to, uh, I'm going to show you guys how I, how I actually use Archon every single day because here's the thing. I am not just working on one issue at a time with Archon. Um, the fix GitHub issue workflow is my most often used, but I'm using it in parallel a lot. So, take a look at this. I'm going to go into my speech to text tool. I'm going to say, okay, I also want to handle more issues. So, let's see here. Uh, let's tackle issue number 167. Um, let's see. Let's also do, um, 182 and 1,087. All right. So, I'm going to send this in. Uh, by the way, usually I'll do even more than this at once, but I just want to make sure that I don't hit my rate limits for Claude right now because it, like, I mean, we're doing a lot of work in parallel here. Not that Archon is token inefficient. In fact, we've been doing a lot of things to make it more token efficient, but still, like, there is a, a bunch that I'm doing at the same time right here. And watch this. Not only can I say, you know, spin up the workflows in parallel, but I can say I want you to, you know, run these workflows in parallel. I want you to wait until all of them are done. So, monitor the workflows until we have pull requests for every single one of them. Then, once we have pull requests for every single one of them, then I also want you to run the validate PR workflow on all these in parallel. And then when it produces the comments and the PRs, I want you to view the pull requests, look at all of the issues that we need to address. I want you to address them and push the changes to the branches for the pull request. And I could even combine all this into a single workflow if I wanted as well. But just like, look at how comprehensive this is. We are going all the way from issue to a final validated pull request and we're doing it in parallel. So, it's spinning up all of these at the exact same time as background processes. So, now if I go into the four shells that I have open here in Cloud Code, we can see that they're, they're all currently running. And then I can go hit enter to view the logs for any one of them. And, um, then of course, I can see them all running in the web UI as well. So, let's go back to the dashboard. And there we go. We got all four of them running. So, we have each of them running in the last 20 seconds here. And then this one that's been going on for five minutes now. Pretty cool. So, yeah, we're still doing the web research on this one here. But, yeah, that, that's how I use Archon on the day-to-day. Like, most of the time when I'm working on a code base, I'm, I'm going to be filing things as issues. Whether it's a bug that I'm going to be working on or it's a new feature that I want to add. Like, both of those, those fit as GitHub issues. And it's also nice because this is kind of like my personal mission control for everything I want to work on. And so, that's why we see like Raasmus opening up so many GitHub issues. Like, we're using this more than, you know, other people are because this is where we document all the things that we want to work on. Like, Raasmus is opening up issue after issue for, you know, bugs and feature requests. And so, that's also really nice because GitHub is where we're going to keep track of the versioning of Archon. Like, as you evolve your code base and you want your coding agent to remember things that you've worked on in the past, you're going to rely on Git commits. I actually love using the Git log as long-term memory for my coding agents. And so, I'm already keeping track of all my work in, in, uh, as like Git commits and so, I might as well like track things in, as in issues as well for, like, the upcoming work that I have. So, you can of course hook in like an MCP server or skill for Archon to use another external platform like Linear or Jira for your task management. I personally just love using GitHub as my task management. So, that's why the GitHub CLI is so crucial for me for pretty much all of my Archon workflows.

Now, of course, there are Archon workflows that don't have anything to do with GitHub. So, like, for example, if I want to create a PRD, I can say, you know, use Archon to walk me through creating a PRD. And so, this is going to, uh, you know, hopefully load the Archon skill here and then walk me through the interactive PRD workflow. So, I'm just showing off another workflow really quickly. And this one's actually pretty cool because it has human in the loop. So, it asks you some questions, it starts building the PRD, and then it stops to ask you more questions. So, maybe we'll see this in action really quick here. And then I also, I do want to get into the other, um, the other thing I wanted to show you guys, like building a workflow based on GSD. Okay. Um, pick your starting point. Let's see. Okay. So, it actually tells me like, here's a few different workflows. Um, I want to use the interactive PRD workflow. So, let's have it kick that off for us here. Okay. Interactive workflows. Uh, got it. The interactive PRD workflow is a guided conversation where AI asks questions and rounds to build out a PRD. Cool. So, um, I don't know. I want to build support for PI agents in Archon. I mean, I think it's kind of an interesting example to let me make sure it actually my speech text tool spells that right. And I'll just say, uh, with the SDK. Not that I'll build this right now, but I'll, I'll show you guys quickly what it looks like to go through a different workflow because like I get so hyper-focused on using the fix GitHub issue workflow. I want to show you guys something a little bit different here. Okay. So, there we go. So, now it kicks off the Archon investigate, or sorry, interactive PRD. So, the workflow is running. It's going to explore the code base before it starts asking us some questions here. And we can obviously go and like view it in the logs along with the other ones. So, this workflow is actually in a pause state now. So, you can see in the web UI, we have support for this where it'll show us like, here is where we have a human in the loop step. So, while the, all the other workflows are running, this one we're more interacting with. So, uh, let me, so it says it's paused. Let me grab the output. So, it's going to read the logs and it's going to ask us some questions here. Right. So, we, the workflow is in a pause state for us to give feedback. And this is very similar to what I was showing in the, uh, example in the diagram earlier where we could have like human in the loop for a planning step, but we get to actually review the plan and have the coding agent iterate on this before we go to the next step. So, asking some foundation questions. Uh, let's see. So, who has this problem? Um, everyone using Archon. I think it's kind of a weird question actually. What problems are they facing? Uh, the problem they're facing right now is there are a lot of people that don't want to use Claude or Claude Code specifically. They want to use different models. PI is more of a general agent that makes it really easy to, uh, to use other models. Um, let's see. Why can't they solve it today? Uh, well, we don't have support for PI yet. And I want to build this now because we're currently working on making big improvements to Archon and PI is one of our priorities. We will know when it's solved. Success. Looks like we can use PI, any model in PI with all of our Archon workflows. So, these questions, um, if you aren't familiar, these are like really standard questions for product managers to ask when they're first creating a PRD. So, we have like a lot of like product manager best practices built into this workflow. Um, now the, some of the questions were a little awkward for me right now just for the demonstration here, but they, they are like legitimately good questions to ask when you're first creating a product requirements document. So, now you can see that, um, using the Archon skill, Archon knows like, okay, let's resume the workflow with an approved state. So, we're approving and then it's giving our feedback in. So, it passed our answers through. So, now the workflow is, it went out of the pause state into the running state. And we can of course view the logs here to see what's going on. So, this is the first node that it ran and then it asked us some questions. Now we move on to the next one. So, it's kind of just like a process of like asking us questions and going in a loop here. All right. Um, oh, is it paused again? Hold on. No, it's still running. Okay. So, we'll wait for it again. Although I might actually not continue with this demonstration because I think you guys get the idea. It really just does this in a loop where it has like a set of questions that it, uh, goes through for each one of the, the prompts that we have right here. So, I think you guys get the point. But there's an example of a workflow where, um, it doesn't really have anything to do with GitHub because not all of them have to. That's what I wanted to show. The final artifact obviously of this workflow is going to be a markdown document, which is our PRD. So, then we would, you know, break that up into tasks and and go through PI loops as a separate Archon workflow to, uh, knock out all the phases that we have in that PRD. And then if you really wanted to get fancy, you could even if you wanted to make it so that the whole like PI loop process and PRD creation, like everything is a single Archon workflow. Like I said, your entire software.

development life cycle, you can package that all up as an Archon workflow. All right, cool. So, yeah, I think with that, I want to show you guys how to create a workflow from scratch because it's actually beautifully easy. So, I'm going to open up another Archon session here. And, you know, before I do that, I do want to spend some more time with Q&A with you guys. So, let me uh go back to the full frame here, and I will open up some more questions.

All right. So, give me a second to kind of read through what we got here and then I think uh yeah, so Raasmus is is here in the live stream now answering questions. Appreciate it a lot, Raasmus. And then thank you Thomas as well. Um, let's see. What is the simplest path to get this to work with linear tasks instead of GitHub issues? Yeah, so my recommendation would be to use either the linear MCP or create a skill to use the linear API and then you would build a custom workflow. So instead of like fix GitHub issue, it would be like fix linear issue or like handle linear task. And you could even have Archon reference the fix GitHub issue workflow and use that as a starting point where you would just mold things to be all linear instead of all GitHub. So it's actually very easy to uh integrate any platform that you want into Archon because for every single node you can inject skills or MCP servers. Um, so for the case of GitHub, like coding agents are so good at using the GitHub CLI that I don’t need a GitHub CLI skill. You might need that for linear though or like I said the MCP server. But seriously, all you have to do, and I’ll show you this in a second, is you like, okay, watch this. Um, I don’t know. I don’t think I’ll actually run this right now, but I Oh, here. Let me go back to my scene so you guys can see my screen. I forgot I wasn’t sharing my screen. So, I can say like, uh, load the Archon skill. I want to create a version of the GitHub fix issue workflow, but specifically for linear instead of GitHub. Like that is all you have to do. Now, obviously, you’re going to probably have to iterate quite a bit on the prompting and like really making things specific to you. Uh, so this is this is certainly an oversimplification, but like this literally could be your starting point. Like this is what you you send in. Um, and so yeah, after a little bit of Q&A here, I’ll show you what it looks like to create a workflow. Like man, the Archon skill is so beautiful because it just it knows everything. It’ll it’ll walk you through everything. And and another thing I love doing with coding agents is I like asking it to ask me questions. And so maybe I would send in, let me go back to my Aqua voice here. So I paste this back in. Um, and I could say like, you know, ask me questions to make sure you understand uh my linear setup and exactly how I want the workflow to function, right? So that way you’re you’re kind of reducing the assumptions that it makes up front. Um, and so what it produces in the end is going to be better aligned with what you actually want before you go into iterating. You’ll probably still have to iterate, but that’s the idea here is like we go through a bit of a planning process with Archon before it um before it creates the workflow for us.

All right, cool. What else do we got here? Um, with the session management within the workflows, how is context persistence handled? Is the data and context passed through the nodes? Okay, that is a really good question and we have a few different ways to handle context persistence here. So, uh, I’m trying to think if I have a good workflow to demonstrate this. Instead of poking around, I might just more answer your question at a high level. Um, but yeah, one of the core things that we have in Archon is a parameter for each node that specifies if we want to continue the session from the prior node or start fresh. And so that that flexibility is pretty powerful because maybe you want a different node where you inject different skills or you just want to like have a new node for some reason because you want to like switch models or something but you want to still continue the same conversation from the previous node. You can do that or you can say like I actually want to start completely from scratch in this node. And then another thing that we have in Archon is we have like outputs like we can output artifacts and so like for in this case we have the planning step and it’s going to output a plan to our artifact directory for our current workflow execution. So this is like one of the primitives we have in workflows. It’s like the artifact dur. And then we when we go in a brand new session in the implement stage, we are going to prompt it to read the plan from the artifact directory. So if I look at the Archon fix issue command, let me open this up here because we you you can prompt inline. I’ve shown that in some of the workflows, you can you can have the prompt right in the YAML for the workflow or if you want to and you want it to like kind of be more organized, you can have it reference a command in the commands folder. So we’ll go to Archon um fix issue. There’s a lot of commands here. So I have to find the right one. There we go. So Archon fix issue. You’ll see here that the argument is the artifact path. So Archon is smart enough to know like once we are done with the planning and we have our plan here, we’re going to prompt in a brand new session for our coding agent to read that plan, right? Like go to the artifact directory and read that plan and then go through the implementation here. So, it’s a brand new session, but then we’re still passing some context or like I said, you can just uh have the context be like continue instead of fresh. So, it’s up to you like flexibility depending on how you typically work.

All right, let’s see what else we got here. Let’s see. Have you really given anything other than Opus a real try at Agenta coding tasks? I haven’t. Uh, so I I do mostly use Opus for my implementation. However, I have tested a lot of using Sonnet for my Archon workflows. So I I talked about this a little bit at the start of the live stream, but I get better results using Sonnet to fix GitHub issues with this workflow than just using Opus by itself in Cloud Code because of all the the context engineering that goes into you know like this. I guess you could call this like a GitHub issue harness like a fixer harness. U so I have run this with Sonnet. So Sonnet is actually the default model for all the nodes here. Uh now for the implementation itself I think we have it set to yeah so we have it set to using Opus so like the default workflow as it stands uses Opus specifically for implementation but obviously if you just delete this it’ll use Sonnet as the default or um you know you can change the model manually or whatever but so like at one point I had it set to Sonnet and I was using it Sonnet for everything Sonnet or Haiku for everything here and I was still getting really good results obviously obviously the best model is always going to give you the best results. Um, but for the sake of like making sure you don’t hit your rate limits and stuff, you can use Sonnet for these workflows. And you can even like ask it to adjust the model at the default level or the specific node level before you run it if you want to.

All right. People like seeing you failing. Pretty sure makes you a normal human. I assume you you referenced that when I was having all of the issues here in the VPS. So, I appreciate that. Uh, we’ll talk about why I have those issues in a second because I have a Claude code. Um, I’m not actually authenticated with Claude. I am authenticated with the Miniax API. That’s what was causing the problems on the VPS because I’m getting things set up for the dark factory.

All right. How many tokens does start the app take? So, when I asked it to just like spin up the front end and back end of Archon, I mean, I don’t know exactly, but it’s probably just going to be a couple hundred, unless it’s like running into an issue for some reason. It’s not going to take that many tokens. And uh, by the way, you can see here that uh, it’s already kicking off the whole process here, helping me create the linear version of the workflow. So, asking me some questions here. I’ll answer those and then it’ll create the workflow and test it for me and everything. Like I won’t go through this whole process here, but for the question earlier on on migrating to linear, like it’s just so easy to have it walk through cuz pretty much like here’s the thing. All of the default workflows we have in Archon, there’s two there’s two uses for them. One is you can just use them directly out of the box if if there is one that like matches how you already work. But the other maybe even like more important part of these default workflows is it’s a reference point for your coding agent to build something that’s actually custom to you. And so even if you’re building something that’s like super different than everything we have here, like some kind of like refactoring workflow for example, um actually we have that as well. So maybe that’s not the best example, but even if you have some like super different workflow that’s like not even close to anything we have here, you still want your coding agent like loaded in the Archon repo to look at these as a reference to understand like the different parameters we have for nodes and how we handle loops or whatever else like deterministic nodes, whatever you might want in your workflow. So very easy to pretty much build any I’ll I’ll show that in a second, but I just want to answer a couple more questions here because you guys have a lot of really good ones.

All right, let’s see. All right. Yep. Pi next week. Yeah, not not a promise that like we’ll have it ready ready next week, but I’m definitely going to be working on me or Rasmus will be working on the pull request for adding in Pi. Um, could I make a node to use codec spark in a workflow? Um, I mean there’s nothing stopping you. You can make a node do literally anything because it can be a prompt to a coding agent or it can be a script. Even if you wanted to use a coding agent that’s not supported by Archon directly, you can just make a script that invokes that like the Yeah, the world’s your oyster here.

All right. Um, let’s see. What about create a node to mention other workflow, not only a command? I I think what you’re asking about here is if we can nest workflows like if we can have a workflow that calls another workflow and uh that is something that we are planning on adding support for. Yep. So I haven’t added that yet like subworkflow execution like you have in N8N. I haven’t added that yet because I haven’t found a a clear use for it for me personally but uh you’re not the first one to ask about it and so like we are definitely interested in adding support for that. Show some love and hit the like. I appreciate it a lot. Yeah, if you guys want to like the stream, I of course would appreciate that. Like the stream and uh and subscribe because I’m gonna be putting out a lot of content on Archon because like yeah, I’m getting pretty deep into the live stream here, but certainly a lot more I can show about like creating custom workflows and things that I’ll be doing with future YouTube content and then of course in the Dynamis community as well. Uh, so yeah, I just want to mention this again. I am planning on doing a lot of of workshops around Archon um in Dynamis coming up here in the next couple of months, including using Archon with your second brain and some strategies to show like how I use my second brain to delegate work across my code bases with Archon. It’s a bit of outside of the scope for our live stream here, but yeah, certainly the second brain is another really big part of of what I’ve been doing and what I’m covering in the Dynamis community. So, if you don’t know, I I did a full 4-hour course on building your own AI second brain in Dynamis. So, I took the entire system that saves me like no exaggeration at least 20 hours a week and I built it like from the ground up in a live stream so that you can follow along. And so, I am turning it into the third course for Dynamis. Um, and so you I’ll put this link in the chat again if if you’re interested in uh really like being on the forefront of Archon and uh going through these workshops and all the courses and building your own second brain. There’s so much value packed into the community, not to mention all of the amazing minds in there sharing ideas every single day. Um, come join us in Dynamis. I I’d love to have you there. Uh, and I I appreciate it. Dynamis community is great. Yeah. Thank you very much for all of you guys who are in the community uh here in the stream. Always appreciate you guys being a part of it. Yeah, appreciate your passion and sharing of knowledge. Yeah, you’re very welcome. It is my pleasure. Yeah, I mean I’m I’m excited. Like we had the question earlier of like why open source Archon? I mean to me like there’s not an alternative. I can’t I can’t imagine building something like this and just like hiding it from the world. Like I just I want people to to see what is possible with AI coding. Like in the end, that’s like my number one goal with my I guess you could call it like my career. Like my with my career in business, like my number one goal is just to uh show people like what is possible with AI and also to just like cut through a lot of the fluff. Like I want to be real but inspirational at the same time. You know what I mean? Like there’s a lot of people that are saying like you can vibe code everything and it’s going to be perfect and you can make millions of dollars. Like I’m not one of those persons that’s people that’s like here’s how to vibe code to make $10,000. Like no, I’m realistic. Like you have to have human in the loop. You can’t just vibe code. Things take time. Coding agents make mistakes. Like I’m real on all of that. But then that’s also why I’m building these tools specifically to add in deterministic steps. Human in the loop. Like building these harnesses because then that allows you to circumvent a lot of these real problems that a lot of people are ignoring. And so, uh, actually going to my YouTube channel here. It’s funny because like, wait, hold on. I think I still have this in my YouTube channel. Um, yeah. Yeah. So, I’ve had this like one-liner for my YouTube channel. Um, for like literally ever since I started my channel in 2024, I’ve always had this sentence um, join me as I push the limits of what’s possible with AI. And I’ve always kept this sentence because that is really what I’m trying to do here. Like in a realistic way. Like don’t you can’t just vibe code a $10,000 a month site in a week. In a realistic way, I am showing what it’s like to push limits of AI and and I really think that like harness engineering is the forefront of AI right now, which is why I’m so excited about Archon. So, yeah, there there’s my little spiel on on on my my vision and and why I want to make sure that Archon is available to everyone.

Cool. All right, John said, "My brain is online thanks to Cole and the community. Couldn't have done it without the crew." I appreciate a lot, John. Yeah, I appreciate you being an active member of the community. Aron just joined the community. Thank you very much. Welcome to Dynamis, my friend. I’m going to be um so yeah, I’m I’m in Dynamis like literally every single day. So, I’ve been prepping for the live stream today. Haven’t gotten to the posts in the last like 12 hours, but uh yeah, I mean like I’m replying to everything like every single day. So, I’ll be sure to to greet you if you want to make an introduction post as well. I’d love to see that.

All right. Let’s see. Any plans to utilize local LLMs? So, that is actually one of the reasons I want to add in Pi because Pi will make it easier to use local LLMs. And then also within Cloud Code, you can integrate it with um things like Olama to use local models. So I I know that so Miniax M2.7 I’m not using that’s not obviously running locally because it’s a massive model. Uh, but this is a demonstration of like I do actually have Claude working with a different model. Like check check this out. If I go back into Claude and I say what model are you like this isn’t just a gimmick. It is actually using Miniax M2.7 through the Miniax API. So you can change Cloud Code to work with other providers like GLM, Miniax, Olama so that you can talk to local models. So I’m adding in Pi because it’ll make it easier and it’s more like natively supported. Uh, hold on, I got a sneeze coming. Excuse me. Um, but yeah, you can you can connect Cloud Code like you can already use Archon with local models if you wanted to try like there’s been a couple people in the Dynamis community that have actually used Gemma 4 with Archon workflows. Like Gemma 4 kind of driving the whole ship. So definitely possible already. Pi will make it even easier though. Yeah. Uh, is there a way to override some settings like the model without changing the default YAML files? So I think there is a parameter to change the default model, but if it’s overridden at the individual node configuration, there’s not a way to change that right now. Uh, but like you shouldn’t be afraid to just ask Archon to change the YAML for you, even if it’s just like a temporary thing where it like revert it after it’s done with the workflow. Um, it can definitely do that. Like I’ve actually been doing that a lot recently with Cloud Code skills where the skill will like change its own scripts as before it runs them and then just like revert it after I’m done with that set of work for any kind of like PowerPoint diagram or PDF generation. I’ve been doing that. So I wouldn’t be afraid to but but still like that maybe that is like a real suggest like not maybe that is a good suggestion to make it even easier. Um, let me actually ask it right here. Uh, so based on the Archon skill what can you tell me about support for changing the model without having to change the YAML itself? Like is there a flag for the CLI to change the model that’s used? So, I’m I guess I’m kind of like validating the answer that I gave you quick here because I mean Archon’s such a massive codebase. It’s hard for me to remember like everything that we’ve decided, especially because like I said, Thomas and Raasmus have been helping me a lot in Archon as well. So, I’m also not the one that’s built everything in the platform. Okay, let’s see. There’s no model CLI flag. Okay, so yeah, I guess I have to take that back. We don’t have a way to do it right now. Uh, you would just have to edit the YAML directly which again like feel free to do that but also like that that is a good suggestion even if you wanted to make a GitHub issue for that in the Archon repo I’d be down to to address that because I’m just going to have it uh okay so I will say that like when I use Archon to work on your guys’s issues in Archon like don’t worry I am actually reviewing your issues and reviewing the pull requests as well so I am like legitimately considering the the things that you guys bring Um, but yeah, just like as the I delegate the coding to the agent itself 100%. We can see that this workflow actually we are almost done here with the GitHub issue fix and then if I look at my clawed rate limits I’m actually curious where we are at with that. Let’s take a look. Okay, it’s actually not too bad. Okay. So, here here’s my claw rate limits for right now. This is this is not too bad. So, um, I’ve used 37% of my 5-hour limit in the past like three hour or four 3 hours and 45 minutes, right? And we’re only at 37%. before. So, there’s a lot of stuff I was doing to prep for the stream today, this morning before I started. So, we were already at like 15 to 20% before I kicked off all these workflows. And we’ve been doing quite a bit with Claude recently. I mean, we have so many workflows running right now and it it it used less than 20% of my 5-hour limit and we’re almost done with all these workflows. Like, they’re pretty token efficient. You could you could run the fix GitHub issue workflow like doz like at least a couple of dozen of times um and and until you hit your five-hour limit and like that’s a lot of work that Claude is doing. Uh, now the really unfortunate thing is my limit reset yesterday and I’m already at 32% for the week. That is a huge bummer. Like I’m probably going to hit my weekly limit around Tuesday or Wednesday. It’s unfortunate. Uh, so yeah, there there are always people that tell me they have like, you know, two, three, four Claude subscriptions. I’m not one of those people, but honestly, after Anthropic u making things making the rate limits more harsh recently, I might have to get a second subscription. I feel bad saying that cuz it’s like Anthropic is being kind of frustrating recently and then I’m just giving them more money when it’s like, well, what can you do? Like I I don’t really want to switch over to Codex because I test Codex from time to time and I I just I think Claude code is better. But anyway, so yeah, it’s yeah, these workflows are pretty token efficient for how much they’re actually doing. Like keep in mind, it’s not just asking Claude code to fix an issue. It’s going through deep investigation, deep research, deep implementation, and deep code review using the right model at the right time. So it’s not uh super tokenheavy.

All right. Uh, working on something similar and yeah, thank you for the $5 donation. I appreciate it a lot. Uh, would love to discuss with you about how we could potentially partner if there’s something you are open to. Yeah. So, Archon is an open-source repo and for me that means that partnerships would be like I’d love to chat and see what your ideas are. But I I wouldn’t really want to like go and work on a separate repo. But if you’re interested in like contributing to Archon and partnering in that way, um I am considering creating sort of like a core maintainer team for Archon. I think that would be the way to partner because I wouldn’t want to like turn this into a separate venture. I like I said with like my core vision and and mission for Archon and my career as a whole is to be open source and share with the world. And so I I don’t want this to like create spin-offs that I’m dedicating time to. Instead of working on Archon as the core open-source repository. But I’m always open to uh collaborations, maybe even I’m thinking like I said making the maintainer team always open to issues and pull requests. And um yeah, sometimes open source can get overwhelming. So also like I don’t always get to issues and pull requests uh even within the same week because it it gets a lot let me tell you guys from all the open source work that I did. But also, it’s more and more realistic to handle everything over time as I have coding agents help me and a maintainer team. And then like I said, like we kind of already have a maintainer team with me, Raasmus, and Thomas. Um, but uh also like thinking of extending that to just like the general, you know, you guys in the YouTube community and in the Dynamis community.

All right, cool. Is it open source available on GitHub? Well, I assume that was a question to someone else because yes, Archon is and and Thomas knows that.

All right. Is it possible to add Gemini? So, the Gemini CLI I don’t think has an SDK. Gemini CLI SDK. So that’s the thing is is for any coding agent that we integrate with Archon, it does need to have an SDK because I I work with the coding agents programmatically in TypeScript instead of using the headless mode for the CLI. That’s like another way that you can automate the usage of um of coding agents. I’ve I’ve been checking on this a lot though because this is another one of the tools that I would obviously want to integrate. A lot of people love using Gemini, especially because Gemini 3 seems to be like the best model for building frontends. Like people always use anti-gravity with Gemini 3 to build frontends and um so I I would love to add support for but they don’t have an SDK. Um, however, you could always use the Pi SDK with Gemini. So we add support for that then it would work. Um, I guess I can’t find an exact link, but um Oh, wait. What’s this? This must not be Hold on. Google GenAI SDK. That’s I think that’s for general agents, not Yeah, that’s not for AI coding. So, that’s not that’s not an SDK for the Gemini CLI, unfortunately.

All right. Uh, to reach these limits, you have to have the product first because you can’t code multiple things without having verticals that do not conflict. That’s true. Yeah. Yeah. When you’re first getting started, it’s like you can’t really do as much work in parallel. It’s more once you have the thing established and you’re just working through different issues, like granular sets of work for improvements or bug fixes, then that’s definitely when you’re going to start hitting the limits 100%.

All right. Um, let’s see. I’m on the $200 max plan. That is true. Would you hit limits much faster on the lower plan? I think yes. I mean, yeah. The answer 100% is yes, cuz the um $20 plan doesn’t take you very far to be honest. And then the $100 plan, like you can you can do quite a bit, but you still hit your rate limits decently quickly. I believe that the $200 month plan is four times better rate limits than the $100 month plan. And then the $200 is 20 times better than the $20 plan. There’s a quite a big difference there.

All right. Let’s see. How is this different than uh BMAD version six? So, okay, here’s the thing. BMAD is a harness. Archon is a harness builder. So, BMAD is a an opinionated approach to building software. It’s a good approach. I think it’s kind of overengineered to be honest, but it’s a it’s a really inspirational approach and you can literally take ideas from BMAD and build it into your own Archon workflows and then customize it more for yourself. So the big thing that I want to be clear on here is that uh Archon is not competing with GitHub specit or BMAD or Cloudflow or GSD. It’s more like those tools are great, but what if you want to build your own? That’s why it’s a harness builder. And so, um, yeah, in a second here, I’ll show you what it looks like to literally just like take inspiration from GSD and like build it as an Archon workflow. Archon is very powerful. Going to scrap the GitLab AI reviewer I developed yesterday and use Archon instead. Archon offers much more room for expansion. Very cool. Yeah, I appreciate it. Uh, and you know what you can do is not scrap what you built, but uh, point the Archon repo to look at what you built for inspiration to turn it into an Archon workflow. And and so yet another thing that I want to, you know, re-emphasize here is you don’t have to ditch what you already have if you want to build Archon workflows because you can bring in your skills and commands. You can reference other other frameworks or other tools that you built to use as inspiration either for the prompting or just like the general process that you’d want to lay out at in a node by node basis. You’re always able to um work with what you already have because I don’t I don’t want to that’s another one of the problems with with all the other harnesses that are out there. Like they’re cool, but they require you to pretty much change how you work fundamentally. And that’s just not going to fly. Especially if you’re working on a team, like if you’re at an enterprise level and you already have a process for your software development life cycle, it’s really really hard for you as a an enterprise level like as a team to adopt something like BMAD because you have to change how you work. But with Archon, you don’t change how you work because you’re building the layer on top of the coding agent that actually enforces that. So you get to even um in a better way take how you work and use coding agents with that. Um, yep Rasmus same here. Use it every day since November. No bans using the Claude agent SDK with the Anthropic subscription. Oh yeah. When are there going to be Archon hoodies and merch? I would be interested. Yeah. Um, yeah. I mean, merch for Dynamis and or Archon would be pretty cool. I don’t know like how much of an interest there would really be for that, but it could be cool. I mean, my man Nate Herk is always rocking his um his merch in his uh YouTube videos now. It would be cool to have a a sweatshirt or sweater or something that I wear when I’m recording. Yeah. All right. Yeah. Very cool. Well, meld Archon and my QA bot. That’s the way to do it. Meld it with what you’ve already got 100%.

All right. At peak, I had 30 something Archon workflows running in parallel across four projects. You know, with the Anthropic rate limits now, you probably won’t be able to do that, but that’s very cool. We we’ve been spoiled at some points. So, uh, within the like, you know, the last couple of weeks up until this week, Archon or not Archon, Anthropic was doing a special where it’s like in off hours you had, uh, two times lower rate limits. And trust me, I was taking advantage of that, chugging through like eight pull requests in parallel for like the entire day straight, multiple days. Well, not the entire day straight, but like during the off hours. Uh, so yeah, it’s it’s a little unfortunate, but I am I am doing some experimentation with like other models that like, you know, Miniax. It’s just not as good as Opus, but you still get quite a bit of power. So, if you have the right harness, you can get really good results even with something like Miniax M2.7. So, I am experimenting with some things that’ll get into the dark factory we’ll talk about in a bit. I am experimenting with some things though to be able to scale the number of workflows I can run in parallel and not have to worry about rate limits. It’ll get a little costly, but I mean the point of using these smaller models is it won’t get too costly.

All right. Yeah, that’s right, Sean. Spicy mango shirts. And man, spicy mango hasn’t come up in a while now. I’ll I’ll need to fit that into more YouTube videos, but it’s like an ongoing joke where um I don’t even remember where it originated. It was like something with GPT where like I was asking for recipes in a YouTube video and it kept like bringing up spicy mango even between conversations where there’s no long-term memory. So, it just kind of became a joke from there.

All right. Let’s see. All right. Um, you know what? So, man, there’s so many good questions in the chat still. But I do want to get to the next part here where I will uh build a workflow with you guys. So, let’s let’s go back into our repository. I’m going to close out of this session and start a new one. All right. So, here, hold on. Let me clear. There we go. All right. So, what I want to show you guys now is how to build a custom workflow in Archon. And there are a million different ways or different kinds of workflows you can build. And like I said, you can take inspiration from all these existing ones, even pointing Archon to look at these to, you know, understand best practices and how we’ve been building workflows already. But as a as an example, like I’ve teased already, I want to build GSD as an Archon workflow. So GSD, it’s a a lightweight and powerful metaprompting context engineering spec driven development system. Little bit of word salad there, but basically it’s a simplistic approach to going from planning all the way to getting your work done with coding agents. It’s it’s it’s nice. And um he I actually really appreciate the inspiration behind the project. So he says other spectrum and development tools exist like BMAD and SpecKit, but they all seem to make things way more complicated than they need to be which I agree with this. I actually already said this in the live stream where these tools are very inspirational, but just like it’s more than you really want in your process and they’re so opinionated. You’ve got your sprint ceremonies and story points and stakeholder syncs and retrospectives and Jira workflows. I’m not a 50-person software company. I don’t want to play enterprise theater. And like preach, man. Like that’s good. So I built GSD. The complexity is in the system, not in your workflow. So the the system gives Claude everything it needs to do the work and verify it, right? Like that’s that’s the harness here is it’s a system for planning, implementing, and validating. A lot of like what I’ve already been talking about. I trust the workflow. It does it just does a good job. It just gets it done. And so I want to take some of the these ideas here. So I’m not like, okay, I’m not going to build this like step by step in Archon. I’m not going to build a replica, but I’m going to take the general flow of the different phases like discussing and planning and executing and some of their strategies for splitting up work into different stages and their verification. I want to take inspiration from this, but I’m going to do it with Claude. So watch this. I’m going to copy this repo. I’m going to paste it in and I’m just going to say load the Archon skill. I want you to help me make an Archon workflow that uh takes very heavy inspiration from GSD, the repo that I gave you the path to here. So, I want you to analyze the repo. I want you to dig deep into the process here and how it goes through the different phases, how it manages tasks and does verification. And I want you to analyze other workflows in Archon, some of the defaults to get an idea for like how those work and how we could translate ideas from GSD into a new workflow that we’ll create that’ll basically go through the same process having human approval gates for different steps of the way within planning and validation. So, and actually one more thing that I want to add to the prompt here is I’ll say I want you to after you do your research, ask me questions. to make sure that we’re aligned on what this workflow does and how we’re going to take inspiration from GSD. And so, I mean, if you wanted to go so far, you could pretty much replicate all of GSD as an Archon workflow. That would take a lot of work. I think that would take too long to do in this live stream, which is why I’m doing it more like and like, you know, let’s just take inspiration from GSD versus replicate it. Exactly. But we’ll still get something pretty similar here. I actually did already test this before the live stream and the results were were pretty good. Now, it is going to take a while to do the research initially or at least a decent amount of time. So, I’ll answer some more questions while we let this run. Uh, but I think this is a nice demo because without getting too complex here, just showing you at a high level how it’s like no matter what you want to build, just describe your process, have it look at existing workflows for reference. And of course, Archon skill helps and then just start building with it. In fact, one of the workflows that we have is a workflow builder. It’s very meta, but whenever we want to build a workflow, we can use this. It just kind of provides some more structure, right? It’s a harness around building more custom workflows. A harness around building harnesses. If I had a dime for every time I said harness in this live stream, I would just go retire.

All right, let’s see. By the way, I built Colm for agent memory context engine. Close to 100 stars already. It’s pretty cool. Congratulations. Yeah, long-term memory for AI coding is uh it’s very powerful, very needed. Let’s see. Um, do you have the possibility to temporarily run commands from the web UI like running a dev server from your work tree to check the look and feel of your change from your work tree? Yeah. So, we have played around with the idea of having like a basic terminal in the web UI to do things like that, but we haven’t built that yet. Um, it’s one of those things where it’s like, let’s see if people really need that before we build it. A lot of what we built into Archon up until this point has been like, you know, just, you know, only like three people really sharing their opinions. And well, okay, obviously it’s not just three people because we’ve had this in the Dynamis community for a while as well, but for the most part, we haven’t had like hundreds of people using it yet or thousands of people using it yet like we are about to. We kind of already have now this week. And so, we’ll see how people really use it. And I’m thinking about um getting like feedback from the YouTube and Dynamis community as a whole for what kind of features we want. Things like this we might want to build, but I just want to be careful with feature bloat. I want to make sure like there are a lot of things that sound good on paper, but like maybe people don’t actually end up needing that. So, for the old version of Archon that I was building over last summer and like released in August, that one we we made the mistake of building in way too many things up front. Like there was a ton of different configuration parameters and stuff that like sounded really good on paper, but then like no one ever actually used it. Like I’d constantly like show things like that on live streams or in Dynamis workshops and people would be like, "Wow, I didn’t even know that was a thing." Like I well I like saw it but I just like glazed over it because I didn’t really care about it. So that that’s the kind of thing I want to look out for. Um, but yeah, I definitely would be interested in like adding more support in the web UI for being able to like actually manage everything yourself, not just relying on Archon. Now the other thing is you could just ask Archon to spin up the site in the work tree, right? Like it it could take care of that even by itself. So you don’t necessarily need a place for like you to run the commands yourself when you can just ask the coding agent to do it. But I I still think that that could definitely be a good addition to the web UI.

All right. Let’s see. How do you pause all or a graceful pause? Yeah, there no way to do that right now in Archon because the pause is reserved for human in the loop workflows. So, it really is like you either abandon the workflow. Well, okay. Actually, there there kind of is support for it. It’s just not direct cuz you can abandon a workflow and then you can uh resume it later. It’s just if you click abandon, it’s not going to like show up in the web UI as something you can resume. But there is a CLI option to resume a workflow. So you could just if you have to like shut it down for whatever reason, you could just say like, hey, like I want you to interrupt and stop these workflows. And then later on you could tell Claude like, okay, resume these workflows. And if it’s in the same session, it’ll it’ll remember the workflow IDs that are stored in the you know the SQLite database or Postgress database under the hood so that it can it can pick up where it left off. So it’ll basically just like retry the node that it that was interrupted and then continue the workflow from there. So yeah, just kind of thing you just ask the agent to do it. It knows how to to how to do that.

All right, cool. So we got we’re only at 50% and it resets in 55 minutes. So our limit is good here and it’s going through. So you can see that it um finished all of the GitHub fix issue workflows. Now it’s just doing the validation here. So if I go back to Archon um man I have so many sessions open up. Uh, which one was it? Yeah, this one here. So let’s see. So it’s checking in on all the workflows over time. Um, done. There we go. So, yeah, you can see that all of the fixed workflows were done. We have the pull requests created and then it launches the validate PR in parallel. So, sorry, I know I’m jumping around a little bit here, but I just want to show you the other thing that we’ve had running this entire time. We created the pull requests, those four in parallel, and now we’re doing the the reviews all in parallel as well. And that’s what we’re seeing in the the web UI here as well. And then we have the interactive PRD that I’m just going to kind of abandon because I don’t want to spend time on that right now.

All right. Oh, yeah, thank you very much. Yeah, congrats on the 200,000 subs. I appreciate it. Yeah, so that happened literally just yesterday. I I reached 200,000 subscribers. So, yeah, pretty big milestone. I don’t get a plaque for that like the 100,000 unfortunately, but it still feels really good to get to that point. So, yeah, thank you.

All right, cool. So, all right, let’s see what we got here. So, we got the core summary of GSD. It implements a spec driven pipeline with these phases, questioning, parallel research, requirements, road map, uh, per phase loop, and then um, we get our complete milestone. So, that’s like the end result of going through the whole spec driven flow. Lists out some key patterns here like the plan checker as a gatekeeper before execution, goal backward verification, which um yeah, I mean that this is the thing with all these frameworks is there’s so much like word soup, word salad, like what does that even mean? And obviously it’s not that hard to like get into the read me and stuff, but I don’t even remember exactly what all these things mean. And I’ve used GSD before. Uh, okay, so here’s what Archon supports for us. The closest existing workflow is the Archon PIV loop which yeah that’s like literally the workflow that I described here in the diagram. So this is the closest to GSD because it has the same stages of planning and then implementing and then validating. Okay, so now it asks us some questions here. Uh, wow, it asks us seven questions. Okay, let me try to get through these really quick. So, uh, is this a full project or a single feature? GSD covers an entire project life cycle. Archon workflows typically target a single feature or issue. Um, wait a second. I’m actually kind of confused by the question. Oh, I see. So, it’s like, do we want to have it build an entire PRD for like an entire project or is it more like a workflow that takes a PRD or feature description?

And runs GSC style plan? Actually, let's do that. Yeah.

So, for question number one, let's do something in between. So workload takes a PRD or a feature description and runs GSD style plan, execute, verify. That makes more sense.

Um, for the planning rigger, yes, I want the full planning rigger of GSD. Uh, I do want all layers of verification like GSD. Where should humans have approval? Let's just do the same as GSD. Actually, a lot of these questions aren't that good because it's just saying like, hey, what part of GSD you want? I really do want all parts of GSD for the sake of the demo here. And, uh, yep. So we'll do parallel research 100%. Uh, where does the progress live? Yes, I want to create similar structured state files as GSD in the artifacts directory. That makes sense.

Uh, naming and positioning. How do you want to position this relative to existing workflows? Something like Archon GSD or Archon rigorous dev? Uh, yeah, let's just call it Archon J-GSD. All right. Okay.

Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly, I wasn't that impressed with the questions it asked me there because mostly it was like, do you really want all these parts of GSD? Which maybe that kind of speaks to how GSD is a little overengineered. Like they claim to be the simple version of things like BMAB, but I still think it has a little much. Um, but also like people get good results with it. Like, trust me, it's popular for a reason. So I, I do want to build an Archon workflow here that, that, uh, really takes inspiration from every part of GSD. So, all right, we'll let it continue to rip here. So it's reading through, uh, some example workflows here, reading through the rule, my rules for workflows and how to build them, gathering context to then create the YAML for me here.

Um, open code integration. Yeah, so open code is one of the, the, uh, coding agents we are, we are considering adding that and PI agents. Yep. We'd love to add more. All right, let's see.

Um, Archon doesn't work on Windows because it can't find the cloud code executable at runtime. Looks like it's a, a known cross-platform issue between Linux, Mac, and Windows builds. Uh, so I'm running on Windows myself. I've never had that problem before. Um, so I'm not sure why that would be the case. If you installed through the binary, there might be a bug in the binary. I would try doing the installation method that's higher up in the readme where you clone the repo and then you go into Claude and you just ask it to help you set up Archon. I would try that instead. That's the only thing I could think of. Otherwise, I'm not sure why it would say it can't find the Claude code executable because when you're running Archon, it's just using Claude under the hood and it's just using it in the same way that you, if you ran Claude from the terminal. So, it shouldn't be different.

Uh, what inspired me to build Archon? Ah, good question. So, yeah, I kind of talked about things related to this in the stream already, but, uh, really it's like I see the direction that we're heading with AI coding. So this is what I talked about at the very, very start of the stream. So maybe you weren't there at the start of the stream, which I mean, probably a lot of you guys watching right now weren't, which is all good. Uh, but like AI started with prompt engineering, generative AI, like that was the big deal. Like, how can we get the single best output from a model? And then that evolved into context engineering, especially for AI coding. It's like creating a whole ecosystem of context for our agent to handle longer running tasks. And that's like the big thing in 2025. And now this year, it's like, how do we create a system that combines coding agent sessions together to do longer work? And how do we, like, really build our coding process as an agentic coding workflow? So the harness is the layer that wraps the coding agent to combine sessions together and add in more control for us. That's how we really get reliable results with coding agents. And so I've been really, really doing a lot of research and deep dives into harness engineering over the past few months. But the problem is we have all of these closed source harnesses like Stripe shared Stripe minions and Shopify's Shopify roast and AWS is building their internal harnesses and and we have things like Ralph loops which are open source but also like not very intricate or like, I mean, the main problem is there's nothing out there that's like custom to you. So I wanted to build a tool that allows you to build your own harness. So no matter what your process looks like for AI coding, you can create it as a full workflow in Archon and then also run it at scale because we support the parallel execution like I've been demoing in the live stream here. So that's the inspiration. It's like I see where AI coding is heading and there's nothing like Archon right now. Like, literally the only examples we have of anything like Archon is is a single harness. So it's not custom to you. It's very opinionated. Most of them are closed source and now this is the layer that build where you can build any of them or build your own. Yeah.

All right. Okay. Honestly,