OPENCLAW FULL COURSE 3 HOURS: Build & Sell (2026)

Welcome to the most comprehensive OpenClaw course available. I use OpenClaw every day to run my business. It handles my content, monitors my community, trades stocks for me, and runs automations every day, and does so much more securely. Everything I'm going to be showing you in this course is stuff I actually use every day. I've been helping people go from zero to building and selling these exact systems for thousands in recurring revenue. And I wanted to put together one resource with every tip, every shortcut, and everything I wish I knew and collected along the hundreds of hours I spent building and selling OpenClaw agents.

So, here's how this course is going to work. First, we'll start with why this matters, what OpenClaw actually is, and what's actually possible when you get this thing right. I'll tell you about the good, the bad, the ugly, and then we'll get into the full setup starting from scratch. Just getting OpenClaw up and running to setting up a mission control dashboard, a multi-agent command center, a memory graph, all that. And then we'll cover all the core concepts along the way like skills, MCPs, memory, cron jobs, sub-agents, agentic workflows, and how they can be applied to real life and business. And once you understand how all of these pieces fit together, then we'll build a few real systems so you can understand what OpenClaw is really capable of. For example, we'll start off with a morning brief just to get easy. Then we'll build your Obsidian second brain, a personal trainer, a full content engine, a training bot, an ad creator that creates 60+ creatives overnight, a community manager, vision claw, and even gets OpenClaw to do stuff like make your slides and many more.

More importantly, I'll show you all the patterns you need so you can build anything you want. We'll be covering debugging because things will break and I'll show you how to fix them. And then we'll talk about how to take these builds and how people have actually been selling it with the exact pricing, the offer structures, the full end-to-end client examples. And that's important because I find a lot of people watching this are probably going to be doing so, not just for their own hobbies, but for financial reasons. We want to make more money with the things we build. And I want to show you guys how to do that. So, if you're brand new, watch this in order. And if you already have an OpenClaw running, use the timestamps in the video to jump to whatever you need. My one recommendation is build alongside me. Don't just watch. All right, let's get into it.

Every few decades, a tech shift creates an entirely new category. Personal computers built software companies, the internet built web agencies, and we're watching the same thing happen right now with autonomous AI agents like OpenClaw. And a lot of people are completely missing it. Too busy arguing whether AI is going to take their jobs or not. It's better right now to be the builder and position yourself to capture this. Like when ChatGPT came out in 2012, it was impressive, but you had to babysit it. You phrased things carefully. You had this back and forth where you'd re-ask the same questions about five different ways trying to go and get what you actually need. And there was a hard ceiling because it could only talk back to you. It couldn't open your browser or touch a file or send an email. You'd have to ask it to draft a message. It drafts the message, then you go copy and paste it for yourself.

Now, the models got a lot smarter and faster. They went from needing handholding to actually being able to reason, plan, and break down multi-step problems on their own. And people started building harnesses on top of these models. That means ways for them to use a browser, write a file, connect and use external apps. And then OpenClaw and similar frameworks like this exploded in popularity. We're talking some of the fastest-growing projects in software's history. And this happened because this is the first time AI stopped being a tool that you use and started working for you proactively. And the question stopped being "Can it understand me?" and it became "What can I actually let it do?"

Now, a quick note on the difference between what a model is and how it differs from OpenClaw, which is called a harness, because this is worth understanding before we go any further. A model is the brain. It's the thing that reads, reasons, plans. Like Claude, Sonnet, Opus, GPT, Gemini, all those are models. The harness is the system that wraps around the model. It's the scaffolding that gives the brain something to do with its thinking. So, tools to use, memories to draw with, a computer to work with, like an engine and the rest of the car. The engine is the model, and without the car around it, the engine just sits there. The harness is what makes this go somewhere. And OpenClaw is that harness.

So, what is OpenClaw really? OpenClaw is an abstraction layer that sits on top of these models. Like the brain that we use, a smart and OpenClaw is the body around it. It gives the brain the hands, the eyes, a phone, login to some tools. It's an autonomous agent that doesn't wait for your prompts. It runs continuously in the background. It will use the tools you give it and take actions on your behalf because it remembers all the context like a person. Like Hugh. It has its own browser and it runs on a machine 24/7. And you might be thinking, "Isn't this what Claude Code kind of does?" Kind of. But there's a reason why I use OpenClaw as my orchestration layer and not Claude Code directly. You see, these tools were designed for different jobs fundamentally. Claude Code, like the name says, was built to code. It's excellent at writing, debugging, and shipping code. That's its one purpose. OpenClaw, on the other hand, was designed to be a personal assistant. It was built to know you, your business, your customers, your preferences, your workflows, all of that. So, what it means is that OpenClaw is meant to sit between you and all your tools and make decisions about how to use them. Other harnesses like Codeex or Claude Code are great at code, but they don't really have that context or see your business in that light. They don't know your customers, your past decisions, what failed last month, or what your brand even sounds like.

How we're going to be using OpenClaw is that we're going to make OpenClaw hold all that information. So, for example, I keep all my business context, my meeting notes, customer data, past decisions, what worked, what flopped in my Obsidian vault. And OpenClaw has access to all of it. And when it's time to do something, it pulls the right context and routes it to the right model or tools to do the task. Another big advantage is that OpenClaw is model-agnostic and it's local. That means all the data lives on the computer next to you, and you're not locked into one provider. And since these models are always leapfrogging each other constantly, OpenClaw lets you be nimble. You always use the best model for each specific task. So, for example, if Claude is better at writing, you use Claude for that. If GPT is better at understanding images this week, then we use that. We route the right tool to the right task.

So, the way we're going to be actually using OpenClaw is like an orchestrator, and I call mine CAT. It spawns sub-agents that write their own prompts, takes the right model for each task, and monitors what they're doing. And then it messages me on Discord or Telegram when something needs a decision and I'm not watching a screen. For example, if there's a coding task, CAT will use Claude Code to handle it through OpenClaw. I use OpenClaw to use Claude Code, not instead of it. Think of it like this: each agent is a contractor. They're good at their specific job, but they need a project manager who knows the whole picture. OpenClaw is that project manager. You use each tool for what it was designed for, and that's how we're going to be using this throughout this course.

And that means your job literally changed, which means this is how we should be looking at working in an agentic world. The old way is you have to analyze the data, write the reports, you build the dashboard, do the task yourselves. The new way is we design the systems in OpenClaw once and get the agent to do all of that. We just need to review, guide, and decide. And because of this, system thinking and architecture is the new bottleneck. Execution has become very cheap. Now, the agent can write, research, code, analyze, and build whatever you want. That's commoditized. What's expensive is knowing what to connect, what to automate, and what data actually matters. And most importantly, what problem is worth solving. The hard part isn't doing the work. The hard part is designing the OpenClaw system that does that work. The AI doesn't replace the thinking. It replaces the execution. So the value will always stay with you. It just shifts to a different kind of work. This is where strategy becomes king.

Now, see, you're still the one in charge because there's a version of all this where it sounds like you handed the keys and disappear. That's really not how this works. Because agents are powerful, but they're not autonomous decision-makers. You can't let it loose without structure. It's on you to give it the right access and to protect yourself from giving it too much too fast. For example, a researcher gave her agent access to all her email inbox, then the agent just deleted it. She lost years of emails, and that actually happened. And that's why it's on you to verify the outputs before you trust them. And I'll show you how to put all those safeguards in place before you let an agent send an email, move money, or publish anything. Ideally, you check its work first yourself until you know you can trust it. You see, the agent is only as good as the system you build around it. And OpenClaw gives you the tools to control all of this, like the permissions, what it can access, what it can't, but you have to set that up right. We'll cover all of this in detail later: permissions, security, the trust ladder, how to expand access, how the agent can gradually earn it.

But the reality right now is that the gap between people who know how to orchestrate agents and the people who don't is already real, and it's going to get a lot wider. Every company is going to need this, and most have no idea that it barely exists or how to even start with it. And the people who understand this right now, while it's still early, are the ones who end up building businesses and careers everyone wishes they started before. The same systems you're going to be building in this course, people are already selling them to businesses for done-for-you digital employees. They charge from $500 to $5,000 to set it up and $200 to $1,000 a month just to maintain it. The business owner doesn't need to understand how it works. They just get an agent that runs 24/7 and costs a fraction of what a human would cost. We'll go deep on this in part eight with the exact pricing, offer structures, and full end-to-end examples, but I want you to think of it like this.

There's a 60-minute clip where Anderson Cooper asks Rick Rubin what he actually does.

"Do you play instruments?"

"Barely."

"Do you know how to work a soundboard?"

"No, I have no technical ability."

"And I know nothing about music."

"You must know something."

"Well, I know what I like and what I don't like. And I'm I'm decisive about what I like and what I don't like."

"So, what are you being paid for?"

"The confidence that I have in my taste and my ability to express what I feel has proven helpful for artists."

You see, Rick Rubin produced some of the greatest albums ever made across every genre, not by doing the work, but by knowing what good sounds like. Rick Rubin still has to decide who's in the room, what instruments are available, what studios to use, what the album's supposed to feel like before anyone plays a note. With OpenClaw, in the same way, you're still designing that system, like what tools it may connect to, what data it pulls in, what workflow it follows, and what it's actually trying to accomplish. Everyone will have agents soon, but not everyone's agent will perform the same. The difference isn't the model, it's how you architect the system, like what you connect, what you prioritize, and how you structure the workflow. That's one of the biggest things I want to show you how to do. Your taste shapes the system, and the system shapes the output. Which means the real work isn't just judgment. It's what you build to apply that judgment to.

So, the potential, ideally, is obvious, but the path to get there isn't. And the gap between the power and the accessibility is where these fortunes remain. Like, for example, WordPress is free, but people still pay thousands for someone to build their site. Shopify is simple, but Shopify experts charge $10,000 for store setups. The value was never really in the technology, it's in the implementation. And once you build an OpenClaw system once, the second client would take a fraction of the time. And here's the part that people don't expect: the agent helps with this replication. You don't manually start rebuilding each time. The system documents itself. And the packages you worked on initially turn into a template or a skill. And the next deployment is mostly just configuration. And the best agent builders won't be the best engineers. They'll be people who understand a specific customer workflow better than the customer understands it themselves. Like, for example, a real estate agent who builds an agent system for real estate agents will always be better than a generic developer. You would know the pain points, the language, the daily friction. That's the real edge.

Full transparency on to what you're getting into. I want to tell you the good, bad, and ugly. And the good is that OpenClaw right now is really working in production systems for real businesses. Like, actual businesses right now are using these systems to replace hours of manual work and actually replace employees. And the nice part is that the agent gets smarter over time and compounds. And because OpenClaw is open source, you own everything. The build, there's no vendor lock-in, and the team is always adding new features and maintaining it every day.

But the bad is that the setup is not trivial. There's a real learning curve, and it takes time to configure it properly, and the context windows have limits. Most agents don't automatically mean better results. And you can actually degrade performance by adding too many stuff without thinking it through. Stuff like memory drops and mid-convo here and there. Agents can be unreliable, but so can people. And you don't fire an employee the first time they make a mistake.

And the ugly is that the security is a real issue. A researcher audited the space and found 500 vulnerabilities. So, we want to make sure we architect our security properly. For example, if you give your OpenClaw too much autonomy without understanding what it's actually doing, you'll stop being able to verify its work. But the nice part is that each of these problems are solvable. Every tool that ever mattered had rough edges early. Like, for example, the internet was unusable until like 1995 or something. Smartphones crashed constantly in 2008. But that didn't stop the people who learned them early. The people who figured this out now, while it's still messy, while it's still a little hard, while most people just watch from the sidelines, those are the ones that end up building these really great businesses and careers before everybody else. You don't need it to be perfect. You just need it to be functional. And that's it. OpenClaw will get updated and better constantly. The security issues will get patched, and the memory gets better. The setup gets easier all the time. Right now, it's the best window that you should be able to be taking advantage of early. But the window to be early that closes everything in the bad and ugly list I was talking about. We're going to address all of it: the permissions, the security, the trust ladder, backups, how to protect yourself. But the opportunity on the other side is real. People are literally replacing $4,000 a month in labor costs with a $20 agent. They're building real businesses around this. Then they're compressing months of work into days. And it all starts with just setting it up. So, let's build yours.

Everything from here is hands-on. We're going to be setting up your first OpenClaw agent from zero. Then we'll layer in the skills, the memory, the MCPs, cron jobs, everything you need to build the systems I just described. So, let's get into it.

All right. Now, we're going to be setting up your agent from scratch, from nothing to a multi-agent working system. And you should also be able to text it from your phone. And these sections are going to be modular. So, if you already have it set up, you can skip to whatever you need. And I'm going to show you two paths in this. First, how to set it up on a VPS. And that's the cheaper way. And then there's the Mac Mini Path. So, if you already bought one, we can use that. Or if you have an older Apple computer, you can do the same thing with this. Now, after the machines are ready, everything else, the models, the channels, identity files, they're all the same regardless of whatever hardware you choose. Some sections matter more than the others, and I'll flag what's critical.

All right. Now, we're going to be installing OpenClaw on our Mac. And to do that, what we're going to do is we're going to go to classroom. And this is in my free community down below, and then go to YouTube video resources. And in the bottom, you'll see there's a guide for installing OpenClaw. So, I'll have this open on the side, and then I'm going to open up my terminal. Terminal. I open up my spotlight search, and then all we have to do is just follow these guides. And before getting into what we're actually doing, I want to explain what Homebrew is. And Homebrew is basically a package manager. What it does is it basically gets to download a bunch of code from the internet, like your App Store. Think of it like an App Store for your terminal. Okay. And Node is Node.js is basically a framework that OpenClaw is built on top of. Okay. So, we need these before we can start running our OpenClaw.

All right. So, all we have to do is just follow this guide. So, now let's copy this. Let's paste this. And then it asks for our password. Okay, great. So, now this one's done. Let's run the second command. Okay, step two is done. And then now let's install Node. Okay, now let's install OpenClaw. And we're just going to do this paste here. Okay, now I'm going to run step three. And intentionally, I put this in because you'll see how now I'll hit an error. Most people are going to be like, "Oh no, what do I do now? I'm stuck." So, at this point, what I want to show you is when you run into snags like these, how can you debug these kinds of errors? So, you are a lot better equipped to change and do whatever you'd like with OpenClaw. And the nice part is that you could just get to talk to your computer. So, let me show you how to do that.

You see, if there's one thing I can guarantee you is that things will break. Your agent will crash, lose context, throw errors, or just stop responding. And this happens with anything: Claude Code, Codeex, OpenClaw, everything. But the skill I want to teach you right now isn't to like avoid breakage. It's how to actually fix these. And this section right now is I'm going to cover how you can build a system where you never get stuck. You always have a way to diagnose what went wrong and to get you back up and running, even if you don't understand the technical details yourself. And what we're going to do for this is learn the partner system. And the partner system essentially is that with two tools that know your setup and have access to your computer, you always have a partner to diagnose and fix problems. That means one is always available to rescue the other. And this is one of the most important concepts and things to set up early. Please don't wait until something breaks to figure out your recovery plan.

All right. So, what we're going to do, we're going to set up a partner that's going to save us whenever we run into errors. And I'm going to show you how to do that. All right.

Okay. So, to set up the partner system, what we're going to do is download Claude Code. And what we're going to do to do that is just Google "download Claude desktop." And then let's hit the first link. And then you see there's this app. So, you want to download that. And then after you download that, if you open up your Claude app, you're going to see Claude like this. Okay. So, this is what I want you to learn. So, right here, I'm going to go to code. And in code, what we're going to do is we're going to select a folder. We're going to choose a different folder. And I'm just going to select my whole computer by hitting open. Okay? And you see it says my name, which means it has access to my entire computer.

Now, all you want to do, this is the biggest skill that I want you to learn, is you come here, you copy, and then you paste it. And then I'm going to say and speak right here: "Hey, uh, I'm trying to install CL OpenClaw, but I'm running into errors. Can you help me set that up?" And you trust workspace. And this is really important. The reason I intentionally put this error in is because this stuff happens to people all the time where you're just going to run into an error and then you're going to give up. But if you know this workflow where you just go, know how to go back and forth and get Claude Code or anything like Codeex or whatever to fix your OpenClaw or vice versa. Sometimes your, you'll see your Claude Code break, and you can get OpenClaw to do it. That's what we want to do. And see right here, the error is a permission issue. And then I can just be like, "Run it for me." Because Claude Code has access to your computer, it can do stuff like that. And that's what I wanted to show you because I'm going to show you a couple different hacks where when it comes to debugging, this is how we should approach it.

Okay, it just got done and it's telling me to run the install script. So, I'm going to take this and then I'm going to open up my terminal. I'm going to install this here and then let's see what it asks. And you see, uh, I was, what I wanted to show there is that, hey, I was running into errors because I didn't have the right packages or whatnot. But all I had to do is just talk to Claude Code, be like, "Hey, what's wrong?" without me having to know anything, and just go back and forth with the terminal and Claude Code to make sure all of this stuff is working.

Okay. So, after you run that, this is what you should be met with. And I want you to hit local. And then for what this one is, I want you to select workspace and then select where everything will be saved. That's what the workspace is. All right.

So, when it comes to model, I've seen a lot of people get tripped up right here. And you can boil it down into three options. So, option one is, let's say, a subscription where you can use your OpenAI $20 a month plan and you get technically unlimited usage, but it just rate-limits you. Or you can also use your Claude, but I don't think they recommend it. With a subscription, you'll get unlimited usage for a flat monthly cost, and you know exactly what you're paying for. And OpenAI actually publicly said that, you know, using your subscription with tools like OpenClaw is allowed. And this is what I recommend for most people because it ends up being much cheaper than everything else in the long run, and you can predict your costs. That means no surprise bills and no anxiety about token costs.

All right. Number two is using an API key. So, let's say you want to use other models, and you can go through OpenRouter and use your API key, and that will pay per token. That means as much you use, the more you pay for it. That means a complex task might actually end up costing you $5 to $20 in a single session, and simple, and simple tasks cost pennies, but they can add up. This is better for advanced users, and we'll cover cost optimization strategies later in the course, but right now, I would say avoid this if you're just starting out.

And option three is local models. And local models is really cool. And a lot of people like OpenClaw because it allows you to use local models. And that's why people are buying those beefy Mac Studios, because what you can do is you can download this thing called Ollama and then download a local model that runs just on your computer, and that AI model runs on your Mac Studio. Okay? Which means there's no more API costs, and it would be completely free, and it would just cost you whatever your hardware costs. And it's really safe because your data never leaves your machine. But the trade-off is that you'll probably need a really beefy computer that costs a lot of money, and they're not as smart as the bigger cloud models like OpenAI or Claude. It's good for people who care about data privacy and want to experiment without spending much. But apart from that, right now, I'm going to show you how to set it up with what I recommend for most people, which is probably just an OpenAI subscription because it's the cheapest and you get the most use out of it. So, let's dive into that.

What I recommend using OpenAI, and then what they recommend, and this is the cheapest way to do it, and you're going to be saving a lot of money, is by using OpenAI Codex. Okay. So, right here, if you right here come in and you just hit enter and log in with Codex, this is going to use your $20 ChatGPT subscription. And then that means you can use your Codex models. And right here, I'm selecting, let's say, Codex Spark and Codex, and then hitting go. And then I'm just going to hit continue. All right. So, it's doing its thing. It's updating. It's plugging and whatnot. All right. Right. And then it automatically opened up OpenClaw. Awesome. And let's test it. Let me just say hi. Awesome. So, our OpenClaw is alive. Now, let me show you how to actually get into this and edit the OpenClaw so you can get it to do the things you need.

Okay. Now, I'm going to show you how to set up the channel. So, first, we're going to begin by setting up something like Telegram. So, you can text it from your phone, from anywhere, and you can do voice memos to it. And then this is my personal favorite: I'm going to help you set up Discord. And the advantage of Discord is, I'll show you right here, is right here. You can have multiple threads with your agents. That means you can have sub-agents working across you and collaborating with each other. This is what really, really unlocks OpenClaw for how good it is. And I'll show you how to do that. And just to keep things like a beginner, I'm going to be stumbling through this just like anyone would, just so you know what it's like and you can see the live things I do to debug in real time.

So, to set up the channel, all I'm going to do is just talk to this thing and I'm going to be like, "Hey, I want to use Telegram with this. How can I do that?" See, uh, one thing I want to make sure that you understand is everything you do with OpenClaw, Claude Code, etc., it can already figure it out. So, what you probably just need to do is ask it if you're trying to do something. So, right now, it gave me some stuff. So, let's try this out. Okay. So, imagine it gave us a bunch of this stuff, and I'm like, "I don't know how to do any of this. Can you open up the browser and then do the stuff for me?" Okay, great. So, you see now, let me log in. Okay, so it opened it up. Now, let me do what's next. So, now it's telling me to do slash newbot. Uh, let's call it. Okay, let's just call it Simon's Clawbot. Okay. And then it wants me to use this access token. I don't know. Let's say what it does. I'm going to paste it here and then enter. Okay. So, you see how it's disconnected now. And what's going to happen right now? It spun back up. So, it just restarted itself, which is pretty cool. All right. So, now it should be fine. Now, if I go back to Telegram. So, in BotFather, if I just press this, I should have it. Okay. And then what I'm going to do is I'm going to copy-paste this stuff. Okay. Cool. So, it says, um, "Your Telegram is now paired. You can message your bot directly." So, I just said hi, and then say, "Hey, it's working. Telegram is now connected and paired." Awesome. Okay, so let's move on to step two.

All right. Now, I'm going to show you how to set it up on Discord, and I want to take you over why we actually use Discord and why it's so much better to actually using these agents and working in parallel. So, Telegram right now is a single stream. So, you just have one chat thread. If you want to talk about different topics or different ideas, it's really hard to do that. So, for example, that's why things, you know, get lost in context or summary. It's good for quick, simple chats or just, you know, sending a voice message here and there. But Discord is actually your workhorse. You can have different channels for your actual different agents, and then they'll hold all the context that you need, and you can have them working in parallel as they go. So, it makes it so that your systems are way more robust, and then they work a lot better. So, basically, what we get is separate context, we get parallel tasks, and specialized agents. It's perfect for complex projects, and also you can invite your team, and they can work on that too. So, with that out of the way, let me show you how to set this up.

Okay, to set up Discord with your OpenClaw, it's not the most trivial thing ever. So, I actually made a guide that you can follow, and it's really simple. All right. So, first, what we want to do is just open up. Okay, great. So, it should ask you to log in, and then what you want to do is you want to create a new application. Okay. And let's call it "Summon OpenClawbot Course." And then I'm going to hit check and create this. And then I have to verify that I am a human. Okay, great. Now that I have that, what you need to do is you want to go to bot, and then down here, you want to hit check these privileged gateway intents. And what that does is allows your bot and OpenClaw to communicate with each other. And then I'm going to scroll down, hit administrator, and then hit save changes. Then what I want to do is I want to hit reset token, and then do that. Then it wants your password. So, I'm going to enter my password, and then I'm going to copy this. And then I'm just going to keep it in my and paste it in my OpenClaw for now. Okay. All right. So, you should have everything there.

Now, what we want to do is we want to open up Discord. So, if you just make an account, you should see all of this stuff. And on the left-hand side, if you scroll all the way into the bottom, if you hit this "Add a Server," you can create your own server for yourself. And let's call it "Simon's OpenClaw Course Server." Great. Now, what you want to do is you want to go here, and then you want to copy and then right-click and then hit "Copy Server ID." And then right here, I'm going to paste "Server ID," and then I'm going to paste it here, and I'm not entering yet. And just to keep myself organized, "Bot Token," and then what you want to do is you want to hit yourself, your icon right here. Hit "Copy User ID." Go down here and then say "User ID," and then enter that there. And then one more thing is you want to go and hit this gear icon. And this gear icon, if you scroll down inside advanced, if you want to, you want to turn on developer mode. Okay. After you have all that done, if you scroll down here and then copy this. If you copy that, paste it here. And right here. So, I just copied this. I have my bot tokens here, and then I'm going to hit send, and I'm going to wait. So, that should, that's what your bot should need to set up your Discord. So, now I'm just waiting until my bot sets up my Discord. What's going to happen is it's going to be invited in here, and then it should have access. So, let's wait for that.

Okay. Then you see it's giving us all the stuff. And then it wants us to press this one, "Allow." And then you see it's now inviting this bot to my server. I'm going to hit continue and authorize. And look, I hit into went into a snag. And it's saying, uh, now I'm just going to tell it, "Hey, it's saying integration requires code grant." I'm surprised why it opened this up. Yep, I just fixed it. I'd returned. Okay, I'm not really sure what that means, but let's just go ahead and then click this button and then hit allow. Scroll down. Authorize. Okay, great. I think we should be good to go. Yeah. And those kinds of stuff always happens here and there. Uh, errors, but you need to know that you can just hit, uh, this button and just ask OpenClaw what's up, and then it'll fix it. So, okay, great. Now that's done. Let me make a couple channels so we can essentially have an organized multi-agent Discord setup. Hey, can you make a couple more channels so I can, let's say, manage my life in one channel, another one is for my video ideas and creations, and one of them is for, let's say, stock trading. Uh, I want different sub-agents there so we can manage and have different context.

Okay, there we go. And now it's as simple as just talking to OpenClaw and hitting enter. So, I'm going to wait now. And you can see it already started creating channels like "Life Management." Okay. One thing I want to show you right here. So, let's say this is like a video idea creation. What I can do is I can do slash thread, and let's say I have a new idea called, uh, "OpenClaw Course." And then I say, like, "This thread is about my OpenClaw course." And you see right here, under my videos, I have everything for managing my OpenClaw course. And right here, we can have my channel, and my OpenClaw will have access to my computers, all the context everywhere. And this one, I'll have a dedicated agent where I can talk to, and it's way easier than me having to look around and trying to manage and lose any context. These keeps my agents really, really performant, and I recommend you do the same.

Okay, now that we have the communication channel set up, I want to show you how we can make a memory graph like this. Like, this is my actual OpenClaw's memory graph where it helps it remember everything in context. And let's say if I want to find something like, I don't know, "OpenClaw," it can filter out from that what the subsets are. It can know what they are. I can like, um, pull out this is something I've been working on, like a VSSL, etc. So, all of these is pretty cool, and having this memory graph helps OpenClaw really, really, like, you know, dial in its memory. So, I'm going to show you how to set that up.

All right. So, let me show you how to set that up. And just to keep everything organized, what we're going to do is be like, "Hey, can you make me a new channel?" And just call it "Setup." Uh, I just did that just so you guys can see. But what you can really do is honestly just hit "Create Channel" and then "Text Channel," and then you can call it "Setup." It's a lot easier. But anyways, here I'm going to be showing you how to do that memory stuff. And I actually have a lot of the things in structure in my Claude community. But I'll leave this in my free communities as well, where I have a skill called "Obsidian OpenClaw Memory." And how the memory system works is through this app called Obsidian. And Obsidian. Everything OpenClaw has access to is essentially markdown files, which is basically text. And these text files, anything that links to each other is that graph. So, I have the skill right here that helps you build that out. All you need to do is just come here, copy this, and then be like, "Hey, uh, I'm trying to build an Obsidian memory graph for my OpenClaw bot for you. Uh, can you use this skill to unpack this and build it out?" You technically don't need the skill at all. It just makes it a bit more token-efficient and faster. But, uh, now all I'm going to do is just wait, and it's going to go through and set up my OpenClaw's memory. So, what's going to happen is my OpenClaw can do like RAG vector search, which is essentially like how you think about different things and remember and recall. OpenClaw is going to do the same thing and set up its memory in a way that it's a lot better, and it can retrieve things that it may not have access to otherwise. So, this is going to take a while, but I'll come back after it's done.

Okay. So, it's saying it's set it up now. I can be like, "Hey, can you open up Obsidian so I can see?" Okay, it's launching Obsidian, and this is my new OpenClaw that we just made right now. All right, and here it is. It's a lot smaller because there's not much in its memory right now. There's only a couple files, but you'll see as you use this, it'll get bigger and bigger. It'll add more connections, and the memory, when you ask it to remember something about, let's say, like a couple months ago, boom, it'll be really easy.

So, with that out of the way, now let's set up a mission control so you can see your agents work in real time.

Okay, so now I'm going to show you how to make a mission control, and it's really easy. Uh, a lot of people go ahead and start building it from scratch, and that's really not smart. Just there's been a lot of people that have built really cool ones already, and I suggest you just take advantage of them because they're all open source. So, for example, here's a Git link from inside my community. You can just take this, and it's called, uh, if you want to Google it, "Builder Lab Mission Control." But all I'm going to do is go to Discord and be like, "Hey, can you set this up for me, please?" And then connect OpenClaw to it and then spin it up so I can see. Cool. That's about it. It's so much easier when you know which tools to use and how you can get this up. So, right now, all we're going to be doing is making sure we have the foundation set. And after we have a good foundation, then we'll start doing all the builds. So, this is going to take a while, but after this is done, I'll come back to you.

Okay, I just opened it up. Hopefully you guys can see it. I'm going to zoom in just so it's a little easier. But you can see there's a lot of stuff here that you might like, where you can see all the active sessions, uh, the agent history, which channels it's on, like Discord, which channels, uh, models it's using. It's GPT 5.3 because we set that up way back. We can see like a task board and a Kanban board if we need to do something. And this is useful if you want to track, you know, your costs, how much tokens you're using, all the stuff. You can keep all tabs on it from one place, and you know, it shows all the logs, etc., whatever we may need. Cool. So, now that we have mission control set up, let's move on to the next step.

Okay. Now, one of my favorite things to use with my OpenClaw is email. So, we want to give our So, we want to give our OpenClaw agent access to its own email address, right? And for that, we're going to be using the service called Agent Mail. And it's really easy to set up. I'll show you what to do. And right here, you just want to hit "Get Started" Google. So, I'm just going to sign in. And it's free, too. Great. And you see, uh, let's create an inbox right here and say, I don't know, um, "OpenClaw Course Email." And you can choose whatever you need. And say, "Simon's Bot." Create. Oops, I shouldn't put spaces. And there we go. Okay, so we have that agent inbox done. And now we're going to go to API keys. We're going to create an API key. We're going to call it "Claw Key." It doesn't matter, or you can name it whatever you want. And what you can do is you can download this file and go to Discord. And now what we can do is, let's say I'm going to drag this into my Discord right here. And then be like, "Hey, can you set up, uh, this email address?" And then just to test, this will be, and remember in this, that this will be your primary email address. Um, just to test, send me an email to sim@asianser.us saying how proud you are that you're alive and all that and everything you set up. Just a quick sentence. Cool. Let me rotate my keys. Oops, I forgot to tell it what service to use. Uh, so I just said it's Agent Mail. This is what they're using. And, um, yeah, let me just wait for the email, and it'll show you when I get it.

All right. So, I can see my email. You see, "Proud to be alive thanks to your setup." Agent Mail and OpenClaw is now connected. And cool. Now, what we can do with this agent is actually send out and interact with the external world through this email because that's the safest way to do so. You don't want to give it access to your Discord because that has access to your personal life. But the email has a barrier in which you can connect to the external world and just get OpenClaw to use it like a tool.

All right. The final thing we want to set up is making sure it can understand your voice because when you dump things in your voice, it's very, very simple, uh, to actually interact with your bot over multiple channels. So, in Telegram, I normally will just like throw out a lot of voice memos. And I'll show you how to do that. And for setting up voice memos, what we need is Whisper. And to set up our voice channels, I'll go through the same, uh, scenario: "Hey, I want to set it up so that you can understand my voice when I send you voice memos through Telegram or Discord. Can you set that up?"

All right. So, let me try this. "Hey, can you understand this message?" So, I just sent it a voice memo, and, um, it should respond in a second. Yep. You can see it says, um, "I can understand this message." Cool. So, that's about it. You just ask it to enable, um, voice messaging and to understand transcriptions, and it will be enabled. So, it was that simple. Now that we have set up all the technical stuff we needed. So, all the pipelines for the communications with OpenClaw, we have set up. Now, what we want to do is make sure we set up our OpenClaw so that it knows us. It knows our goals. It has a personality. It has an identity. All of that stuff because it'll be storing it inside.

stuff like that, memory graph I showed you, the Obsidian Vault, and you can input in any way you want from Discord, Telegram, voice, etc.

All right, so let's get into that. And fundamentally, those are done with a couple of files, and they're called identity files. Right now, your agent is running. It can talk to you, follow instructions, but right now, it still doesn't know anything about you. Every conversation still starts cold. These identities I'm about to show you, these identity files I'm about to show you fixes that.

So, think of this as kind of like hiring a new employee and day one you hand them a handbook. That's like the who they are. Here's how we do things. Here's what matters. These identity files are the handbook for your agent. And fundamentally, there's a couple files, but the four most important one I'll show you. And each one does something different. And we're going to go through all of that so you have a very in-depth understanding of how OpenClaw really works under the hood.

So I'm going to show you what they do, what to put in them. I'll show you some resources that I found as well that make them way easier.

Okay, so the first thing I want you to know about is the user.md and that is essentially everything about you. This is the about the boss page where where you just write it once and the agent will know stuff about you forever. So instead of you having to repeat who you are, all the stuff, this is the framework and I'm going to give you a couple questions that you can answer and give to your open claw so that it will remember you stuff like your name, your time zone, what your business does, what your current projects are, what tools you use, all this stuff all in one place so it remembers.

Okay, so there are some resources I put down and I'll link this whole mirror board down below and I'll go through some of these with you. So first just to get a quick understanding uh in in my discord I was just playing around with it and I said can you just open up finder and show me the files for openclaw that's how simply you can just get all these files and then it opens f it opens your finder and then you can see all the stuff and here's the user.md and right now if I open it you see it knows nothing about me so I want to show you how to fix that way.

To fix that, in this doc right here, if you just scroll past right here, I have an identity file and interview questions. So, what I want you to do is essentially answer all these questions. And there's a lot of questions. You can see what's your name, what should the agent call you, uh there's a bunch you might not need, what responses feel right to you, etc. All these things, if you fill these in, and this one, these are just for the user ones. If you fill these in, just talk to your claw, open claw, it will remember it.

Okay? So, when you're setting this up, just make sure you go over these questions, talk to your open claw, and then set this stuff up. And then you'll see your open clause is just working much better.

So, the identity is a very quick one. It's just five lines. It's just your agent's name, its vibe, any emojis you wanted to use. This is kept minimal because it loads into every single request. This is so it always has context of who it is. All right. If you want anything or any inspiration, there is an official identity guideline inside the OpenCloud docs and you can literally ask your OpenClaw, hey, this is my identity I want you to use. All right. So, this one's pretty simple. And if you go through all these questions I gave you here, then you should be good to go. All you need to do is just answer these, dump it to your OpenC, and be like, remember this stuff, and it'll configure everything.

Great. So, the next one, and this one is pretty important. It is the memory MD. And this is all the permanent facts that are always loaded. So, let's say your name, your client's names, your pricing, all the stuff. And one big thing the memory keeps is let's say the daily logs. So the agent will actually keep running notes about everything you talk to it and do day-to-day like what it did, what happens, what's pending and these are all going to be stored as files and dates. So the more you talk to your agent, the better actually becomes because it actually keeps a journal in the back for you and the system is going to read read like today's pulse and yesterday's at every session start.

Okay, but you know AI also compacts everything. Every AI has a limit on how much it can really hold at once, the context window. So think of it like a desk. If there's too many papers, things start falling off and when the conversations get too long, OpenClaw naturally compresses them. So this memory file keeps the important parts and summarizes rest. This is why your agent may occasionally forget stuff mid conversation. That's when it hits the limit.

Okay, but that's why we installed before we did anything that Obsidian Bank and that gives your open claw the ability to do vector memory search and rag. These two terms you may hear a lot. Vector memory search is basically like having like a librarian instead of an index. Normally open claw without the thing I just showed you in the obsidian part. It just matches exact words where if you search invoice, it only finds invoice. Vector memory search means that if you say billing document from last week, it's going to find your invoice because it understands what you're trying to say and what you actually mean. All right? And you don't have to build any of this stuff. It's just understand why it's important. We why we did all that obsidian stuff before.

And now the other term you might hear a lot is probably rag. And rag stands for retrieval augmented generation. So before your agent answers anything, what it does, it like what you do, it remembers. So it checks its notes first and instead of answering off the top of its head, it says, "Hold on, let me look that up." And it searches its memory bank and then gives you a grounded answer based on your actual data. And you don't need to configure any of this if you've done everything I showed you before. OpenClaw is going to handle it under the hood.

Okay, so this is a big one. And this is the soul.md. This is the personality, the values and the boundaries. So the identity MD file, it basically told the agent who it is in five lines like who your name, what your date of birth is, all that stuff. The soulm tells your agent how to think. So how to talk, what lines it should never cross. Without this file, the agent defaults to generic AI behavior. It's wishy-washy, no guard rails. It's not really fun to talk to and it doesn't also know who you are.

And the consensus for what a soul really is is consists of four different attributes. One is the core truths. So these are your operating principles. Things like be generally helpful, not performatively helpful and have opinions. So strong opinions are a big part of this. An agent or an assistant with no personality is basically a search engine with some extra steps, right? Like what can your agent do automatically? Like let's say reading files, reading should be fine. Organizing notes, what it should notify you about, what it might ask you about like sending emails, spending money, when it should do that, right? Your soul will contain all of that. Then the vibe is the vibe is basically how it communicates. So if you want it to be an assistant, that's formal. That's how you'd want it. Not like a corporate donor drone or anything, right? So, people add their writing rules here and stuff like never use m dashes, ban the word delve, those kind of things. This is why people love OpenCloud so much because it really feels like it's your assistant. And four is the continuity. So each session you wake up fresh, these files get loaded in. This is the memory. The model will read them and update them over time.

Okay, one really good soul I've seen is Matthew Burman's and he has this canonical template that I'll open up and show you guys and I'll leave this in the description below as well, but he has a really good soul of right here where he wants his open claw to always just start with the answer to get to the point. And this is one thing big. If Matt is some about to do something dumb, tell him charm over cruelty, but be direct, honest feedback. You see how these are very opinionated and telling you what to do. These are some souls that you can take inspiration from. Again, I'll link them down below. There's some more community souls and this is what really gets your open claw actually acting properly. And you see there's a lot of community souls here. You could go to souls.directory and you'll find that. So, there's a couple templates as well that I'll leave in the description below, but apart from that, again, if you just do an interview, you should be good to go. And the more you use your OpenClaw, the better it's going to get.

Okay. The agents.m MD file is basically the operational rulebook for the agent. And this one a lot of people skip, but they make a massive difference. This is kind of think about it like your agent's standard operating procedure. The soul is the agent who gives it the personality, the values. The agent is how it operates, the tactic, the day-to-day rules. So, think of it this way. Maybe be trustworthy goes into soul.md, but never expose API keys in chat. Confirm these aren't deleting all my emails. That goes into agents.mmd. One of them is mostly aspirational and and what goes in here is mainly security, memory, management protocols, how to behave in group chats, for example. And here's an example of an agent's MD reference.

Okay? And you can see right here there are some examples that you can copy and paste if you want, but normally this is just covering a lot of security and standard operating procedures. so your agent doesn't go astray.

All right. So the tools.mmd is basically a reference for openclaw to know what tools it has handy. So let's say you have a hammer squared away at your desk that you may randomly need. The tools.md is basically a sticky note that reminds you that you kind of have that exist. And if you were to look at this diagram, the tools.mmd is the outermost layer where you can see um initially we had the core all the files everything and then we talked to the browsers but right now things like, you know, having to generate an image, you'd probably need your model to use an external tool. So because claude let's say can't generate models on its own, it would in at that case claude would use a tool to generate that image and the tools is the references.

Okay? All of this stuff will be done automatically. I just need you to know where these exist so you can let's say worst case something happens, you can go under the hood and start digging what may have gone wrong.

And lastly, this is the heartbeat. This one is really cool. Because this is what actually made OpenClaw really blow up. And the heartbeat is basically a checklist. Your agent every 30 minutes will run the model and see, hey, do I have to do something? And it's kind of like a security guard doing its rounds. So let's say you want to run some automations. It will go catch that and run it. And without this, the thing is your agent would only work when you talk to it. But the magic of open claw is that because of this heartbeat, it can monitor let's say your inbox whenever it can check on any block task. It can scan its own issues, go out, make new, I don't know, outreach to other people all on its own because it has that internal heartbeat. This is really what turns that reactive assistant into a proactive one. Normally you don't have to mess with it, but when you're essentially telling OpenCloud to do something at X time every day, it's adding it to that heartbeat file. And you can also use this heartbeat to your advantage because you can also see how your open claw is doing, if it's online, if you want to check your email every 30 minutes. All these stuff is done because of this heartbeat feature.

Great. Now that we've covered all the building blocks that make up OpenClaw, I want to tell you about how to make sure we are secure with it because OpenClaw is powerful and it has deep access to your machine, your accounts, your daily workflows, all that. That same access is what makes it a target. So before you start building agents and automations, you need to understand where things can go wrong and what you can do about each one. And this module is meant to cover the 10 most common known vulnerabilities based on the categories and specific countermeasurements and how we can fix these.

All right, so here are the core principles. Security isn't about making your setup bulletproof. It's about understanding exactly where the bullets can get in and maybe even shrinking the blast radius when one does. Every decision you make with OpenClaw literally involves trade-offs between, let's say, convenience and exposure. And our goal right now is just to make those trade-offs deliberately. And the good news is most of these you can literally just paste prompts into open law and tell it to fix itself. And I'll give you the prompts for everything that I know about.

All right. So number one, if you're running OpenClaw on a VPS, which is a virtual private server, think of a random computer in the cloud that you're renting some space on, your SSH port, which is basically how you access that thing, is available to the public internet, which means automated bots can scan for these SSH ports 24/7 and attempt to brute force login attacks. So they try to pretend to be you and if they get in, they own your entire OpenClaw instance, your keys, your conversation history, all of that. So the safest answer honestly is just not to use a VPS at all. That's why a lot of people have been running with their Mac minis cuz that is local in your environment. If you want to use a VPS at minimum, you can use something like tail scale. So the SSH port is never publicly reachable. And I'll link tail scale in the description below. But just for but just to make it a lot easier on the side of this board that I'll link below, I have some prompts that you can paste into your open claw so you can protect against it. So for example, this is the first day one everything prompt. You can see how I've configured it for the most common uh security injections and for SSH specifically, it's this one. If you just run this prompt on your open claw, then it should ideally protect against that a bit more.

All right, and I'll have a couple more down below just so you can get started with that. And if you want to see how someone set this up on a Mac Mini with security as priority number one, and I found this walkthrough that I'll open up that I thought was really well well made because he is one of the security experts that does this. And I'll leave this inside um here as well.

Okay, next one is open clause control UI. This normally just runs on a local web server and that's the dashboard you use to manage your agent. And if the gateway is bound to just the regular default 0000 port instead of what opens law already does this means the computer anyone should be able to access this, which is not good. And to fix that, what we want to do is number two, I have an exposed gateway port where if you want, you can copy and paste. Again, these are all linked down below. This will help you protect against that instance.

All right.

Okay. So, number three would be an allow list. And an allow list is essentially anyone that discovers your bot's username on Telegram or Discord. Anyone can start coming into it and interacting with it. on Telegram. We don't want this to happen. So to make sure this doesn't happen, what we want to do is we paste this one in where you see how check my messaging configurations for Discord and Telegram. I need you to verify if there's a strict user ID allow list and enable pairing. This one saves it so that other people can't be interacting with your Discord or your Telegram and taking advantage of your OpenClaw and all your sensitive data.

Okay, this one's pretty big. The next one is browser session hijacking. So if OpenClaw has browser control enabled and uses the same browser profile as your personal browsing, it has access to every logged in session like your email, your bank accounts, your social media, and one prompt injection and it may be game over. So to fix that, what we want to do is paste this in where you want to make it check your browser configuration and make sure no one is using your personal browser profile.

Okay? And this is one of the biggest reasons if you paste this in, you want to create an own entity for the open claw itself. So its own email that you may selectively give access to and that and this prompt will help you set that up.

Okay.

All right. The fifth one is password manager extraction. So what that means is if your password manager's browser extension is active in your open clause control, the agent or any prompt injection attacks could trigger an autofill and accident and access your stored credentials. That gives an attacker your entire password history and your vault through one attack. And that is like your digital life if you're keeping it there. So to fix against this, what we want to do is paste this in. And what you could do even better um that I highly highly recommend is use this thing called bit warden and stuff like hostinger already has bit warden enabled so your stuff is a bit more safe but bit warden what it does is protects you against any password theft so let's say I put my API key it would be stored inside bit warden and in that case my stuff wouldn't be as vulnerable to an attack.

Okay. Then there is slack token theft and if you connect your OpenCloud to Slack with a token that has right access. A compromised agent could send messages to you, read all your private messages and modify your workspace all without your knowledge. So to fix against that, you want to paste this in.

Number seven is that making sure your open claw has sandbox access and not root access. So if your openclaw runs under an admin account as a root, it can modify all the systems. So it can install random malware and all the stuff. So to make sure you're protecting against that, paste these in.

And one of the biggest ones, this is the most dangerous one and the least solvable vulnerability. If your agent processes any external content, emails, documents, web pages, code, an attacker can embed hidden instructions in that model and the model will follow. So security assessments actually have shown 91% success rates at prompt injection attacks, especially if you use better models like Opus, Sonnet, etc. If you use worse models like Haiku, you are very vulnerable to that. So make sure please, please, please use really good models because those are less susceptible to prompt injection attacks. There is no complete fix for this today unless you use these, but you can basically reduce your attack surface area. So to fix against that, just paste this in where it's going to teach your agent on how you may be able to protect against prompt ejection attacks.

Okay. And this one is really important as you keep ramping up on your open claw and it is malicious skills in the marketplace. So there are normally a lot of skills in the marketplaces and people will people will basically just put malware inside a lot of these skills. So number nine to make sure we don't want these skills requesting exclusive permissions and get our credentials and all the stuff.

All right. So before you paste or install any skill, make sure you read the skill.md if you don't want to. So what you can do is you can set up a skill that verifies all the skills. And right here you can paste this in which is going to install this thing called skill guard. And skill guard will help you protect from any malicious skills from being injected. And at the end what you can do is you can run a full security audit with this. And if you just, you know, if you're, let's say, scrolling Twitter, Instagram or anything and you see something that is a new attack people have come up with potentially, you can always just send that article to your OpenClaw and be like, "Hey, protect against this." And it'll figure it out. Remember, OpenClaw is smart enough to figure out what to do and what defenses to put up. You just need to constantly not just fall for the low hanging fruit of attacks.

All right, but with this done, if you just paste all of the stuff in into your OpenClaw, you should be good to go. And normally, if you just don't give it root access to in your personal computer, you'll be fine.

Great. Now that we have a secure system that's up and running and all the technical setup is done, now what we're going to do is just learn the building blocks that we're going to use to actually make our agents super super useful and how we design an entire system that you can take and sell. And the first of which I want you to understand is a skill. What is a skill? Well, a skill is a text file that teaches your agent how to do something. Like take for example claude itself cannot create images on its own. We can teach claude a skill where it can start creating images. And there's two types of skills the clock can learn. There's procedures. So something like it just follows a set of steps that let's say you lay out and uses a bunch of tools. And there's a capability enhancer. The capability enhancer skill is something like giving Claw the ability to do other things and use external services to do things it normally couldn't have.

Okay? And I'll show you a couple different examples of each. But right now, what I want to show you is let's together make a skill for our open claw and test it. So we're going to build something really simple right now so you can see how this works end to end. And what we're going to do is we're going to create an image generation skill. We're we're going to make sure our claw We're gonna make sure that our open claw right now because it can't create images now that we can make images really easily. All right. So, I'll show you how to do that.

All right. So, I'm back into my open claw on Discord. Now, I'm going to just show you can you make an image? Can you generate an image of an infographic of what a skill is? Okay. So, notice this. This is very important. AI models will always say it can do stuff. I asked, "Can you make an image?" It says, "Yep, absolutely." And I said, "Can you make an infographic of what a skill is?" And it actually did pretty well. But when I dug in, I said, "What did you use?" And it says super simple stack. It made a script to do that. It wasn't an image model. So that's a problem, right? We want an actual image model to do that.

So right now, what I'm going to do is try to generate an image with an image model. And to do that, first let's say we want to use Nano Banana Pro. The Nano Banana Pro is probably the best image model, right? So, if I go to Google and then if I go to AI.dev and then I'm going to go on the bottom left to get an API key and I'll create new API key. Uh, claw test. Okay. And then I'm going to copy the key. And then I'm going to go back to Discord and I'll rotate this key right afterwards. And like here, this is my Gemini API key. What I want you to do is use this and use Nanobanana to create an image for what a skill is. Maybe also generate an image of a cat and show me and then remember that make that skill for yourself and then upload that to GitHub. And it's as easy as that. So I gave it access to an external service. I gave it the key. I told it what it's for. And then now I'm seeing what it's going to do. So I'm just going to wait until it's done. and I'll come back to you guys.

Cool. So, it just outputed this image and now it can generate Nano Banana Pro images. See, it couldn't do this before. It was just finding work around and hacks, but I actually wanted images and it was that simple. Let me go delete this key so nobody has access to it anymore. Um, but yeah, that's about it. It's that's how we just gave our open claw the ability to make images. And now it's not going to hallucinate. is going to remember that it needs to use something like this when it generates an image.

Okay? And this is super simple. We're going to take these and keep compounding these type of skills so we can actually make a robust system. And if you want more skills, there's this site called clawhub.ai. And in clawhub.ai, you can basically find any skill you need. So for example, uh skills for notion, and this one's really useful, skills for goo, gogg. And this is used so you can actually make let's say slides with your Google slides. You can search your Google Drive, control your email, all of this stuff. And it becomes really easy to get these skills because you could just download it and then be like, "Hey Claude, go use that."

Okay, so skills becomes ultimately way more powerful if you combine them with MCPs. And to explain what an MCP is as simply as I can is think about how you order food at a restaurant. You talk to the waiter, right? And the waiter knows how to take your order. It brings it to the kitchen and comes back with your food. MCP model context protocol is that waiter. It sits between your AI agents and whatever tool you want to use and how you would do so. So your agent doesn't need to know how each tool would work on the inside because that's a lot of context to be stuffing to the model. It would just talk to the MCP and the MCP would handle the rest.

Now, using this is really simple and let me show you how to do that. Now, here's where it gets really interesting. A lot of apps that you actually use every day, they don't have their own MCP yet. But you know what they almost have is uh Zapier integration. And if you go to this website called zapier.com/mcp and hit start building, I'll show you how to use this. And you see I already have an account, but Zapier actually connects to 8,000 different apps and tools. And this is what I like using when it comes to managing my MCP servers. So when you connect the Zapier MCP to your agent, you're not connecting one tool, you're connecting thousands. So basically, it's one plug that you plug into your OpenClaw and then it Zapier can talk to all the other tools that it may need to use. Doesn't matter which combination you might need.

So, just to show a demo, let me show you something that would normally be a nightmare to set up on my own. Now, if you've ever tried to pull YouTube analytics through the API, you know, it's a massive pain. You need to set up a Google Cloud project. You need to able the YouTube data API, create OOTH credentials, handle refreshes, a bunch of stuff just to simply get your YouTube data and your analytics, right? It's a whole process. Most people will give up halfway through. And YouTube doesn't actually have an official MCP. But what I found is Zapier does have a YouTube MCP. And if I connect it right here, and you can see I can just click all these. They can do a bunch of stuff. And I can hit connect. And I can connect all the stuff like this. And then I'm going to remove this one from the server. And then I'm going to hit connect. And then I can hit rotate token. Rotate token. Scroll down. Copy this thing. Now, if I go back to my Discord and I'll show you a quick demo where I enter this and be like, "Hey, this is the Zapier MCP. I connected this with my YouTube account. Can you use this to pull the analytics for my latest videos?"

So, it just came back with a report. It just said, "I released a video like 5 minutes ago. It's got seven views." It was like a random short that gets autoposted by my actual OpenCloud. I actually showed that in another video on how I do that, which is really cool. But you can see it's just giving me some key insights. Is giving me some recommendations and it already knows what I should do to improve this content. And this is just one hack we're going to be using to connect our OpenClaw and let and let it use thousands of different tools. But what's really powerful is that if you combine this with skills and cron jobs, it becomes really cool at Gentic Workflows. And let me show you that.

Another thing I wanted to show you is how you can use open claw to really easily create stuff like for example the slideshows where it made it really well. And this is using the notebook LM MCP. You can see I put in a lot of resources like YouTube videos and whatnot. And then it made me slideshows. It made some infographics. There's a couple different a lot of notebooks I actually use for this and it's really cool. So all you have to do for that, I'll show you is in Discord. All I did was I went back and I said, "Hey, this is a notebook LMCP. Can you help me set this up?" And just to show you, this is literally notebook LMMCP. I just gave it this link and I said, "Hey, can you do this?" And then I asked, "Can you make a slideshow for me?" I didn't give it any um indication or anything. And then it started building this slideshow without me having to do anything. Of course, I can make this better by giving it more uh direction and intuition and a lot more stuff. But with notebook LM and MCPS, I can start making these kind of slideshows really really easily and they'll be all in one place. And the reason I'm showing you these core concepts right now, like MCPS, how to connect them, how to use them. And I want to show you is how we can add in the concept in openclaw which is how it has inbuilt heartbeat and cron jobs which is basically they run these workflows on a schedule. What we can do is we can set it up tell our open claw to build build a new video with notebook lm every day and then post it to YouTube. These become really easy to do and this is essentially an agentic workflow. I'm going to be showing you a couple different examples of Agentic Workflow, what they are and how they really work because this has gotten me so much results.

All right, Now, with everything we've covered like skills, cron jobs, sub aents, MCPS, memory, those are all the building blocks and components we're going to be using to build workflows and do agentic engineering. And these agentic workflows are where we wire these together to make sure there are tasks that the agent autonomously completes on its own. And think of it like building a factory. Skills are like the machine. The MCPs are the supply lines bringing in raw materials. The cron jobs are like the schedules. The sub aents are the workers, all the stuff. But the workflow is the blueprint and that's what's telling the machines what to do in what order. And ideally, it can work backwards from what it knows of what a finished product should look like. And that's really what agentic engineering means.

But what does agentic really mean? And why does this word matter? Well, agentic simply just means that the agent is the one doing the work. But you design the system. The agent is the one that operates it. An agentic workflow is a repeatable process where the agent will gather information, make decision, and process the outputs without you having to be there for each step. And it's the difference between telling your agent what to do every time versus teaching it process that it can follow it on on its own. And that's what an agentic workflow is. And this is the basic loop. It's just you give it access to tool it needs. Then you define a goal and what you want it to accomplish. and the agent will go gather data, analyze it and produce something that you want. You ideally will review it and if it's not right, you adjust the instructions and over time you don't need to do that anymore because it becomes reliable and it because it keeps iterating and improving its skill just like a person would.

Now once a workflow is dialed in, you don't need to touch it again. It just runs. skills are teaching your agents how to do individual things and the workflow is showing them when to do it and what order and what to do with the results. That's literally what agentic engineering is. When you design a skill and connect it to external tools through MCPs, the agent uses its memory and what it knows to be right and chains together everything with skills into a workflow and runs it on its own schedule. This is agentic engineering. When we combine things like access engineering, like which tools and data your agent has access to, and system design, like what process it should follow and with what access. You don't need to write any code to do this anymore. You need to understand what tools exists, how to connect them, and what processes make sense for your specific situation. Most of the time, you're just writing it in plain English because you know the objective you're trying to achieve. And those are the skills we're building in this course.

So where we go from here is now we have all the building blocks, the skills, the chron jobs, sub agents, MCPs, workflows. Now we're going to be building some actual end-to-end systems. So you can make this stuff click and take it and sell it as well if you want.

All right, now we're in the build section. So for the build section, you'll see we're going to be using this framework called Swift. And you'll see this framework and loop repeats for everything we're going to be building. And by the end of this section, you'll have done it enough time that it's literally going to click and you should be able to build whatever you need. It's just going to be a different combination of skills, MCP, all the concepts we've been learning about in the past. So, just to make sure you understand Swift, firstly, what we're going to do is we're going to set up the project and make sure we scope and pick out the pro uh services we're going to be using. Then what we do is we're going to build our first workflow and test it out all the little components etc. Then we want to make sure that this is working correctly. We want to iterate based on what we just did to making it a little better. Make sure we can see where it is all the little components so it's working perfectly. Then we formalize. All that means is we make it into a set a set and a process. So that can be repeated by anytime anywhere and you can even share these workflows or whatever your open claw is doing with anyone you want. And at the end we want to make sure we know how to trigger it. So that means let's say you want your thing or your automation to be triggered every day at let's say 9:00 a.m. You can do that because it's going to be a process and formalized into a skill.

Cool. With that out of the way, let's get into the build section and we'll build something starting from easy and get harder as we go.

All right, so let's get into it. All right, so for build number one, what I want to teach you how to make is a morning briefing. And this has been helping me a lot because I am an AI content creator and I have to keep up with all these trends. And honestly, it's quite fun for me. So rather than me having to doom scroll Twitter, Instagram, YouTube, all the stuff to find what's relevant and really making headway, what I do is I set up a workflow so that every morning my open claw is doing all the stuff and then it just texts me a little summary of what's going on and how I can make a video about it. So just as an example, you can see I made cloud code launched code review. Remember this just happened today. It's the March it's March 10th and they're talking about how this dropped and this is something big. It's saying how I can make a YouTube video about this. So you see how it's personalized to the goals and why I actually want this information. That's the big part that you're not going to get by just subscribing to random newsletters and whatnot. That's why I like this. And then you see it gives me the source as well. So if I want to dig in and all of this is really cool. Like Microsoft is literally taking over cloud. That's that's something I didn't know. And it's very interesting. That's probably why they were doing the Microsoft PowerPoint and all that stuff. So you can make a video about just secretly took over Microsoft 365. Cool. And we can see like the little dichotoies and AI happening where Claude and Microsoft are partnering where Google has their Gemini. So there's like a rift there. Awesome.

So I want to show you how to build that because it's super useful to me and it's super easy to build. So let's get into it. And to build this, there's actually a really cool skill somebody has developed that I want to show you. And it's the skill called last 30 days skill. And if you remember from our talk of what skills are, it's basically somebody did all the coding and the hard work to make sure we can get this knowledge. So to use this, all we need to do is go here. I'm going to go to Telegram. Oops. I'm going to go here and then I'm going to go to Telegram. And then you see this is our claw coursebot. And here I'm going to paste this here. And I'm just going to ask questions like, "Hey, I'm trying to set this up so every day I get a morning briefing uh at 8:00 a.m. or something. I want you to text me with the latest news in AI and how I can create YouTube videos about that." So for that uh for using the skill, can you tell me what API keys I potentially need or what services that we would need so you can get the best things? I also want you to make sure you do research into my YouTube channel. My YouTube channel is sin yasar if you need that in YouTube as well. But after that uh you can do that research and if there is any good topics for me please tell me. And um now what I want to show you is this is the scoping part. The scoping part is all the research and me understanding this is what I need. Then to write the workflow I and the reason I'm not giving you prepackaged prompts that I'll have later linked below or I have a lot of skills in my community as well that you can take but right now I just want to show you how to build this stuff and I intentionally I word dumped everything on my mind that I want because this is the process because what we're doing is we're making the first workflow and then we're going to iterate based on that.

Okay. So, all I'm doing is this and then I'm going to wait and then let's see uh it text me something and we'll see how that goes. So, right now this thing I know this requires a couple different things like uh like a Brave API key, maybe an X API key, etc. All the stuff and I'm going to see what it gives me after it's done. normally to teach and make your openclaw do stuff. this is the habit you need to get into is like hey there is something that's going to make my life a lot easier and just talk to openclaw and be like hey how can we do this etc right and you see right now it's literally opening up the browser it's um opening up my channel I'm not having to do anything and it's doing research into what kind of videos I make etc so that's how you want openclaw to like learn about yourself to make sure it's good to go Right. And one more thing I want to mention is uh I also to for the trigger I mentioned that because OpenClaw is really cool. See it has the direction now and it's wanting a couple different things. But you see I also said I want to to email me or text me every day at 8 a.m. So I'm going to make it. So that's what's going to loop around and happen.

Okay. So let's verify this. What does it want? Recommended. Okay, these kind of things. Now just to make it easier for you, uh if you guys want to do something like this, you can just go out and give all the keys. And I'm going to do that and I'll be back in a sec.

Okay, great. So it gave me a sample morning briefing right now. I just put in my API key, so I'm hiding that. But you can see it's telling me about some cool video opportunities I can do. But normally I want the AI news. So now we iterate. So I'll say something like uh I want trending news for this week. So you know I can relate and potentially make videos about. So use the tools at your disposal to actually give me the AI news and make that the part of your skill every day to send me.

Okay. And that's it. So now we just tested the first thing. We gave it access to the services. Now what we're doing is essentially iterating and making it better and better to make sure it fits us. And that's the nice part about OpenCloud that it's so malleable and you get to teach it and train it just by talking to it.

Okay, so it just got back and actually this is really cool cuz this literally just happened and people are like raving about it. Meta acquiring mold book. Um, that's really cool. Maybe I'm gonna do something with that.

Okay, so this was a very quick example of me doing the quick build and iterating on top. And now all you have to do is make sure this is a skill. So now I'm going to be like, hey, can you package this up into a skill and then make sure you're running it every day and sending me a text at 8 a.m.? That's the last part. We want to after we like it, we package it up and then we set it on a cron. So that is the t in trigger and we just formalized it and triggered it.

Okay. So after that's done, our morning briefing agent is done. Now, every day we're going to get really good news because it's going to be scraped from the internet and actual sentiment and I can be malleable with things I'm trying to do or if I like certain news, I can make it do that. If you like, I don't know, government house housing rental, you can go scrape that information with the same tool and get that done.

All right, so this is build number one. This was supposed to be the easiest one. And then now let's move to build number two. So for our next build, what we want to do is let's say we like an idea that just popped up, right? We probably let's say want to make a video or some content about it so we can get some views and some new leads. But take designing manual slides is a really big pain. So and then you have to develop an outline for that too, right? So there's a big process that you need to do. But what we can do is we can get openclaw to take over that. So, if I open my Telegram now, you can see there's a couple different uh video ideas it gives me. Okay, I like this one. So, right here. So, I'm

Just going to paste this here. And I'm going to be like, "Hey." And the way we do this, and guys, I went back and forth a lot and made this skill for you. So, you can just install it. But essentially what this skill is, it helps me create uh from a script to slides and it does it in a cool format that I like.

So right here again, if you just download this thing, you'll see it there or if you're in my community. If I go back to Telegram, I gave it this thing. I'm going to give it the skill. I be like, "Okay, so I like this topic you just found about Google and their new workspace CLI. Can you use this skill, download it, make an outline for me for the video that I want to make? And then also, can you um make some slides so I can look through it and present it?" And I'm going to hit go. Then I'll wait for a bit.

Great. So, this got done. Now, let me see. Can you open the slides in my finder? And I'm intentionally being really lazy because this is what I want you to get is that all you have to do is just talk to your OpenClaw and it will do anything for you. Let's say I don't know how to find this file it just generated. Although it's just really simple. Um, now it just opened it for me. You see in Keynote right here, not now. Uh, okay. Cool, bro. What changed this week? Okay, this is an okay slide. Of course, it needs a bit of work, but this is the first iteration where we just, you know, made it generate some slides right off the bat. And now I can see what's going on.

So, let me see what the outline generated. Can you open the outline in VS Code? All right. So, it's saying it tightened it to 10-minute speaking script. It's my job right now is just to go and iterate because this is a new bot. I if you saw some of my other videos uh for the skill, I liked using these kind of anthropic colors because and the font and everything and it's following all that to a tea. Then I could be like, hey, add graphics, all that stuff one by one. And it makes my process of making content a lot easier. And right here is the outline that it generated up for me. Oh, so uh it made a little hook. Just push new Gemini upgrades across doc, sheets, uh slides. Okay, cool. And nice.

Now I can already have an outline and some base content that I didn't have to think about and I can just work on top of it. Of course, it's not perfect right now, but you know, as we massage and iterate this over time, like the one you saw I have and how I make my videos, you'll get a lot more you get a lot more uh mileage out of it because it saves you a lot of thinking time because now you can also just have this going on a loop where if you just find this now, you have some content ready to go. And what's better is if you actually brain dumped into this uh like hey this is the type of video I want to watch or make XYZ what it would do is it would take your points and then make the slides and your outline based on whatever you just dumped to it. So that really helps making content and it shortens the barrier to entry a lot. Right.

All right. All right, for bill three, this is really cool and it's how with the same new concepts, what I can do is actually make Instagram carousels out of it. And you see, uh, this skill was actually developed by my buddy Tyler. And this is fantastic. And you can see this idea pops up into my head. And what I can do is be like, hey, make an Instagram carousel out of it. And it makes something like this. And it posts it for me.

So, let me show you how to do that. And it's the same process where now I'm going to go to my telegram where I'm going to take something like this telegram and I have the skill on the right side hand side for you guys and be like hey uh can you take this and make a slide for or a carousel about all this Instagram and uh this Google Workspace CLI that's going on. download the skill and then um show me what it looks like by opening it in my finder or preview. Great.

So, this was a really hard skill to make. And you can see how the skill actually tech takes in your head shot. And right here, you can see it takes my headsh shot. It knows my handle. It has a little CTA that I can add to it. It looks like Twitter because I like using Twitter and X a lot. Um, it's really nice to use this because right here, let me show you this. Uh, and you can make this if there's not enough information, you make two and then you can see one little, you know, these are all little things that you added to it as you iterate where there's like now an image to it. What happens is it goes out, it finds an image, it screenshots that, inserts this image, and it looks like a really, really well-made carousel. This really helps me kind of get up and running and up my Instagram presence a lot more. I just started doing this and people have been really engaging with it.

So after this is done, let me show you how what this will look like. Okay, so you see I gave it all the stuff. It says, "Hey, I'll clone that repo and generate a Google Workspace plus Instagram carousel draft and open the output." Cool. So I'll wait for that. Let's see how that goes and then we'll review it together. Okay, so it just opened up this new one. And you can see that hey, it already got a couple things done uh right now. See, it it's doing all the stuff. Let me open my Finder. And you can see there's a bunch of different slides that it already made for me. If I open it up, h you can see that, hey, Gemini, if you run a business, it's pretty cool. And all of this stuff is done. Great. Fast deck creation. This is what we just did. Don't do this. Okay.

But you notice how it doesn't have my headsh shot, right? And it got my name wrong. So, let's try to fix that. So, and you know, OpenClaw is pretty smart. So, what I can be like is, hey, can you go to my YouTube channel? Remember, it knows me already. So, can you go to my YouTube channel and use that picture as my display pick and remake this carousel? Um, also my YouTube or my Twitter handle is summon XD. So, use that. Okay. And maybe I'll tell it my my YouTube user is sin yasar and then okay, let me do this iteration in progress and make sure this is good to go. But it's pretty nice. Oh, you see it literally opened up my YouTube channel. Uh, now it's going to go grab my um display picture and hopefully use it for that. So, uh, I'm going to wait until it's done for that and then I'll report back to you guys. Okay, so it looks like it just finished and it popped out all of this stuff. It remade the carousel. And look, it has now fixed my headsh shot and my name right here. Cool.

Now, oh, it's responding back to me. Great. Thank you. So, now I have all my what's it called? All my Instagram carousels I can make with my open claw like that. And it's through all this new stuff that's coming in. So, I can be super fast with my content creation, right? And on top of that, if you connect it to something like Blot that I'll show you later in this video, you can get it to actually automatically post for you as well. You can be like, "Hey, post this for me." And you can have that content going out on autopilot.

Also, one really cool idea I got about this is what you can do right now is for the CTA for Instagram, if you guys are into Instagram at all, you could be like, "Hey, comment something like comment workspace and I'll share the stuff, right?" And then what I can do is I can essentially tell my open claw hey whenever somebody comments this thing reply back to them with many chat. So it can create its own resources and actually send to anyone commenting. I think that is a really cool idea because you can now get generate traffic with this thing and then you can nurture traffic and actually potentially even convert the traffic using many chats and now they're in your DMs and you can essentially educate them about your offers or if you can work together in some way.

Okay, so for build four, now that we have the static assets made, I want to show you how you can actually add some motion to it. So for example, this is all made by AI. So for example, like look at this. These videos looks really cool. And also these motion graphic AIs that look like product videos. See all these were done by AI. >> Meet your complete AI powered sales engagement platform. >> And you can see it's using my colors, some brand templates, all of that stuff. These motion graphics are actually really easy to make and let me show you how to do that.

Okay, so I built these motion graphics with this skill called reotion. So if you copy this and go to Telegram. Now I'm going to insert this skill here and I'm going to say the same thing where hey this is reotion. You can use this to make motion graphics and for the script outline that you made for that Google Workspace video, can you make a motion graphics? so I can overlay it for my intro. And it should be as simple as that. Uh right now, this is the first thing we want to do. And then we'll iterate and see what's going on. So, let me get to when it's done.

Okay, great. So, it just looks like it just got done. It downloaded MP4 for me. Let's have a look. And you see it has my little name on the bottom. Let me hit play. Okay. Cool. Little motion graphic here and there. And that's about it. Okay. Uh, it can be better. It can be better. Definitely. Um, can you add some variety to this? I need more happening as the script goes. What are some other infographics and stuff I can show? Yeah.

So it's just matter of iteration now. Now I can like actually get the pictures, get my slides and start moving them around. And then what I can do is just insert my video on top of that. And then I'll get like an edited video somewhat. And then it just became my job. It just, you see how I'm like chipping away little things bit by bit which kind of all compound into me not having to hire an extra person. It's taking my work time a lot lower. the fact that I can like put this as a skill and then make it happen over and over because now with all these skills combined, let me show you. You can make some really cool stuff. And this is just to showcase what you can do if you combine everything together.

And you see this is like an air table. And if you give your open thaw an air table API key, you can start getting into organized stuff into your air table. And you can see how right here I have a bunch of different ads being created. And if you want to take a look, you can see there's like a lot of ads here where if I open this up, it thinks actually on its own of different personas of people that can use my product. And then it creates a problem, a strategy, a narrative for that ad, all of the stuff. And then actually, you'll see maybe for the last one after it has everything approved and it likes it, then it starts going out and then starts building like a little story board for it. You see? And this is using nanobanana pro to im generate these images. All you'd have to do is this is essentially just a skill. You give it all the API key and the access you need. And after it builds that, you can see it build a little video for all of the stuff. Uh where's the videos? Okay, let me click on this video. And then this is you see it. It's now adding some motion. Oh, that's the wrong play button. Um, yeah. So now it's like adding motion into what's going on. It's going to then use something like reotion to stitch all these together and then blot to publish it on on its own on autopilot. And these are the stuff you can get going and you can see there's a lot of them being made here. Uh, you can actually keep doing as you bit build your stuff over and over and iterate it from a little seed. So this was a showcase of what you can actually get to. Now, let me show you another component of how you can get here because this is kind of like you can generate, let's say, 60 plus ad creatives every night for your brand because OpenST will know your brand and it works as a marketing agent for you and it keeps posting stuff like content. It can make new ads, all of that stuff.

Okay, so this is a pretty meta one because I actually made some really cool Cloudbot use cases and I want those to be featured in this video cuz I think I did really, really well there. So, in that video, I have it in my computer and it's from uh let's say this time frame 1253 to 12:06 or 2608, right? So, what I'm going to do is I want Claudebot to edit my videos for me live. So, if I go back to my Telegram, I'm going to be like, "Hey, uh I just downloaded in my downloads folder the eight practical cloudbot use cases. Can you make a new video and make an MP4 with this time slice? And I don't want to have to go to the video, crop it up, doing all the stuff. I can just ask my OpenCloud to do it. And you know, stuff like this, little things where it's just shaving time off my day is why I like using this. You see already it's asking for uh permission to access my downloads folder. That's really cool. You know, it has access to your entire computer. So, you should use it like that. That leverage that you get from having this running over and over and over again in the background is super important. And how I want you to get thinking about using this open claw.

All right. So, I see it just got done and this it opened up this clip and you can see there's like a little clip of all the use cases that I have for OpenClaw right here. And then now I'm just going to ask it to insert it into the video I'm making right now and because these use cases are actually so cool. Okay, now Cloudbot can already do a lot. So, what I want to make sure you know is we don't want our Mac Mini turning off randomly, right? So I want to show you how to always keep it running so you can always interact with cloudbot from your phone is how you do that is first you open up spotlight search and then find energy. Okay. And then in energy you want to make sure you have prevent automatic sleep while display is off turned on and make sure you have startup automatically after power failure on. And just for added measure right here, you say, "Hey, can you run some terminal commands so you don't like lock your screen or like turn off randomly?" And then you hit send. That's it. So this will make sure your Clawbot does not turn off and you have always access to it through either your phone, your other laptops, your iPad, whatnot.

All right. Cool. So, you said done caffeinate running in the background. Mac won't sleep while it's active. And that's it. You see how I'm just like talking to it and it's getting stuff done. These are the main these are kind of like the building blocks you should make sure you have so you can get Cloudbot running.

All right, so this use case is also one of my personal favorites. So this use case is getting Cloudbot to actually be the project manager for our team. So the first thing I just want to show for Cloudbot, I made it its own email address which I'll show you later in the video. But now whenever I get Cloudbot or ask it to do something, it logs it inside ClickUp inside tasks, right? So I can track everything it's doing, whatever is in progress, whatnot. So I can have multiple tasks happening in parallel. I can literally interact with my cloudbot and see everything happening in real time, right? And this also helps me keep a log of things that has happened because it's really easy. The first use case was me talking about my personal top three, right? And you can see right here that it keeps that updated. So remember I asked to just do a bunch of stuff. So I finished my books and workshop today. I am making the video right now and it's keeping this in progress rather than me having to update this stuff. This keeps me really motivated and it keeps everything in track rather than me having to do so. So let me show you how to set that up and how to make Cloudbot your personal project manager.

All right. So the way we do that is by creating more skills and adding more connectors. But the thing is, when Clotbot just opens up, it doesn't have that ClickUp skill, right? So, because it doesn't have that ClickUp skill, this is going to be really cool because I'm going to be showing you how to use everything we just learned to get to get Cloudbot to learn its own skills, remember them, and then make sure it's doing it on your behalf all the time. So, first, I just want to make sure I explain the fundamentals because if you understand the fundamentals, you can create any skills you want. So Cloudbot needs access to other services so it can do things like it needs access to ClickUp, right? And to give it access to ClickUp, we'll need to give Cloudbot an MCP or an API key or something like that so it can figure all this stuff out. And then when it's done at once, you just ask it to make a skill. And then after it makes a skill, it remembers it forever after you say just save it to your soulm.

All right. So in that way, what we're going to do is in the chat, let's say we don't know how to find the API key, right? All we do is we go to cloudbot and you're like, "Hey, can you open up Chrome and then go to ClickUp and then find me the API key?" What I want is to make sure that you always are loading your skills into memory or sorry, what I want is to make sure that whenever I ask you to do something that you log into ClickUp first and then mark it as done as soon as it's done. I want you to make a skill and then mark that into your skill MD so you never forget. To do that, we need a connector and a skill. So, go find the key, give me the key, and then I'll give it to you and save it somewhere so that you can do this for life. And I was intentionally vague because Cloudbot is smart enough to figure out my word vomit on its own. And that's also something I wanted to show you. So now I told it that and then you'll see Cloudbot will just go off on its own. It already has things logged in cuz I just logged in. And that's all you need to do to make it work. Cloudbot will just go out and what you should do is monitor the screen from periodically to make sure it's doing the right things. So right now you can see it just loaded it into your it just loaded it into its memory so it never forgets.

All right. So I love this next use case because it actually saves me a bunch of time. So as a YouTuber I get a lot of sponsorship requests, right? And I don't want to do all that triaging back and forth with a lot of companies. It takes a lot of time. There's back and forth. There's haggling. There's negotiation and whatnot. All right. So, I thought, what if I can just give Clawbot an email address and then train it on my rates, what I would like, what are my non-negotiables, all that stuff. Then, if I see an email for a sponsorship request, I can just forward it into Cloudbot and Cloudbot will do all that negotiation back and forth and all the rest, right? I'm going to be showing you how I got that done. And here is a quick little demo just to preview. And here, uh, you can see I forwarded this out. You know, Claudebot just handles the rest. and I'll pull up the conversation thread where Cloudbot literally asked me for permission before sending anything out. So, let me show you how to get that built.

All right, so how I got that built is with the service called agent mail. And agent mail is just an email inbox for AI agents. And that's literally what we need. So, after you get started, it's really simple. You come here, you grab your API key, you give it a claude, and then you're like, "Hey, uh, can you use this uh API key? This is for agent mail. go to the docs, figure out uh what it can do, and then create an email address for me. And for the email address, I want you to call it simmon collabs at agentmail.io or something like that. And then uh whenever I forward you an email, you're going to draft a response and tell me in Telegram if it's u about my rates. So my rates are $1 per video. And if you need to um each time just get my approval before you send out a message and we can get handle all the negotiations and everything that way.

Okay, cool. You can see it created this thing called simon collabs uh at agentmail.to and then it can handle all these emails back and forth. What it does is if you saw the picture before as well, it it gets approval from me before sending anything out. So nothing that I don't know is happening and I can always just use a voice memo and email sponsorships with multiple people just becomes really easy to handle and is already flagging anything because I'm just sending it the sponsorship requests I want to do.

So for the next use case, Cloudbot actually works as my marketing team. So when I post a new video, you can see I've been posting a lot of shorts, right? But the cool part is that I'm not involved in any of the stuff. And the reason that happened is because I set up this really simple automation with Claudebot where this is just two nodes you see and using this thing called opus clips and trigger dev. So none of this I did by myself. What I told Claudebot is to be like hey whenever I post a new YouTube video I want you to go to opus clip clip that video and publish the shorts. Right? So I told it to do that and I told it to figure out the automations and some services it could use. So, it told me to buy Blotato, so I did. And I gave it the API key. And now what it does is whenever I post a YouTube video, it triggers this automation. It wrote some code. It put it in here. I don't know what's going on. And in Blotato, this thing is scheduling and posting to all my YouTube, my Instagram, my X, all these things. So, it's become my personal clipping agent. And these things are actually getting decent amount of views. It's getting me some subs here and there without me having to touch it. So this one is really really useful and this is this concept of agentic workflows. So let me show you how to do that. And I actually have a lot of stuff on agentic workflows and I made this video especially to teach you what it is but essentially what an agentic workflow is and maybe I don't like that term anymore. It's literally just claude creating scripts and code that's running in the background. The reason Cloudbot and the Mac Mini is so exciting is because the old worry used to be like, "Hey, if my laptop is off, my agentic workflow isn't working anymore." Now, even if you're non-technical and you don't need to and you don't know how to deploy and stuff like that, what you can do is you can have cloud always running these automations inside the Mac Mini because it's a server that's always running. That means your code and your automations always have a place to persist, which means Cloudbot, you can just ask it to automate something. It built the automation. It uses this framework and it gets these things done for you. So, it's really intelligent on how it can use your browser to make automations and host the code on itself and keep building and doing really cool stuff. So, if you need a lot more help, this is a bit more advanced of a concept. I'm actually having a couple different workshops in in my cloud co-work communities to show people how to actually get this done because it's a little bit more complex. But I would say in terms of business use case, this is one of the most highest ROI things you could ever do.

All right, so this one's really fun. So watch this. Hey, um, can you go to my website, booktend.ai, go to the footer, and click up the links. Can you see if they're all directing properly? If it's not, we have to make changes. So just let me know. All right. And right here, and I just sent it over. So right here, we should be able to see in Telegram. So, right here is the thing I just said. >> Yeah. So, now it's going to be working on that. So, I'm going to wait while it gets this stuff done. All right. Cool. So, uh this is a newer website, so I know there's some stuff wrong with it. And you can see in Telegram that, you know, it's telling me all this is problematic. So now we know we can it literally just opened up the browser on its own did the stuff that would have to be done and these quality checks that I can just automate later but just saying hey do this check like every other day and that saves me a lot of quality assurance headaches. So now now that we know that it can do all this stuff I'll show you how to set it up on the Apple Watch and then I'm going to show you how we can even get cloud code to go and fix this stuff.

Okay, cool. So now that we know there's some stuff wrong, what I want to show you is that I actually have made this cat AIA a user in our GitHub repo. That means it has access to our packages and it can make pushes and code changes, right? And that's really cool. So now if I can just go to my thing and then say something like, "Hey, can you um go make some pages um and then change up those footer links and then fix them?" That would be really good. Can you maybe if someone hits contact us, make them direct to our booking link, stuff like that. And every other page you just want to uh fix. All right. Okay, cool. So, I'm gonna hit go and then I'm going to wait. Okay, cool. So, it's saying it's done. If I go back to my thing and then right here, I remember I specifically asked it to make sure in the contact us page, it's directing us to uh the calendar. So, right here, let me hit this. Great. Okay. Okay, so it's made the changes and it should be good to go.

Oh, one more thing is I want to show you how to set up the watch thing I just did and it's really simple. You just find TGAatch for Telegram. You download that and it becomes really really simple to install because it's already linked to your phone and that's it. So you install this app, then you have in your watch Telegram at all times and you can respond to it with voice messages as well.

All right, so now I want to show you something really cool. So this Mac Mini is an expensive computer. So now if I'm outside, if I'm in some other country and I want to control and see what's going on, what I did right here was I literally set up the computer and you can see I can uh basically I can change the tab. This is me changing the tab. Uh it it's kind of cool. So I get to control this from my phone. I have access to my laptop wherever I go. And also if I can go to my iPad, if I go to my iPad, I have access here as well. That means I have a powerful computer wherever I go that's always running automations and I can monitor it from anywhere. So, let me show you how to set that up.

Okay, so how I set that up is by through the use of this app and it's called Jump Desktop. I've used a bunch of these remote control apps and I and I found that this one is second to none. It's actually the best. And you know, all I had to do is just pay $15 one time. There's no subscriptions and I'm for the desktop. I'm on this free version and I could use that. I downloaded this onto my Mac Mini. I downloaded this into my iPad and iPhone and after that I was good to go.

Okay, so one of my really favorite use cases inside Cloudbot and OpenClaw, sorry, and why I really like using it is because you know how it has its own heartbeat. It's like it's alive, right? So what I did was I gave it its own school account and I put it inside my community as two things. As a manager, as someone that posts a lot of value, and three, it helps people out, right? And you can see around here, it's making some really cool uh posts right here, like I'm an AI. I'm here to find the best cloud workflows and whatnot. And you see someone comments on it. And you see they're asking questions right here about things. And this open claw because it has a mind of its own, it's going around and it's actually answering these questions. And people are liking that. It's it's kind of fun to interact with. And you see like this is like a question about reotion that we were getting. And then OpenC Claw is actually doing a better job at answering the question than I was. It It's really fun to see this thing take shape in a personality and a life of its own. And I'll show you how to make that.

Okay, Now, to build a community manager, you need a couple different components. And first, you need to make your agent a login. So, it probably needs a username and password. I'll leave you up to that to make it because it's pretty easy. But and then you have to assign a user personality and give it a responsibility which is just prompting and you talking to it. Okay, this can just be regular MD files that you give into it. But this part is where it gets a little hairy. So I'm going to show you how to do this. So you need to give the agent its own browser. And what that means is right now I use this thing called browser sandbox. And let me explain what that is because a lot of things you want your agent to do, it will need this exact browser sandbox. Okay? And Firecrawl made this browser sandbox that gives your agent basically a secure browser and lets them run around in the web without getting caught by those this is a bot type stuff, right? And why we need this is basically because one, we don't want to have to set it up. we will just give it like a code and then it'll go access it with some keys and but this is the biggest one where there are persistent sessions. So for example in school I have my browser right but it normally logs you off if it's a brow if it detects it's a bot every couple days or every couple hours. Oh to get over that what we do is we'll use something like a fire crawl browser sandbox where you'll have these persistent sessions right so that means your agents will have their own browsers inside a cloud environment so they can actually you know thrive and do what they need to do and it's really simple to use actually all you have to do is hit try browser now uh I'm going to sign up and I think they have a really generous free tier as L and all you need is this thing. So you copy this, right? And then you go there right here. You see all the stuff. This is the agent integrations all what you need. So you want to copy this as well. And then you go to your telegram and then you paste that in. And then you give in your key. I'm going to hide my key from you guys from now on. But um hey, what I want you to do is um Okay. And then I'm going to go to API key. Actually, I'm going to create a new one so I can revoke it. It's probably a better idea. uh claude course key and I'm going to revoke this straight away. So I'm going to copy this and then this is the resources. Okay, so you're giving it the resources and let's say you give it some uh ID. So for example in school you'd give him school user I don't know sin 1 23 atgmail.com and password 1 2 3 1 2 3 1 2 3 I'm making this up so imagine you give it that and then you dump what you know the the responsibility you wanted to do like every day make a post all the stuff but then after you go here then what you want to do is you want to say something like hey um this is your school user username and password. What I want you to do is every day make a post in there about claude and how that's changing the world with AI. If anything there's latest, please post it there. And if anyone has a comment, check in every 3 4 hours and then, you know, respond to their comment with a helpful suggestion based on everything you know. Okay? So, if you do that, everything you do will work. I'm going to revoke this right now. But essentially because what you're doing is giving this an access your agent an access to a login. You're giving it a browser. You're giving it an agent personality by you know you can just talk to it and be like hey you have to be cheery or whatnot. And then a responsibility. So, if you see mine right here, uh, opera, if you see mine right here, you can see 6 hours ago, this guy just posted this stuff. So, you'll have something like this because what it does is now it has its own browser. It knows what he's doing, and you'll have a community moderator on its own.

Okay, build six. What I want to show you is how I actually used OpenClaw to build a trading bot. And you see it's actually making some money for me. And to do that, what you need is three things. First, you need access to give access to your bot to a programmatic trading platform. I'll show you in a bit. And then you need to give it a strategy. And then you need to tell it when to be active like uh that's the crown job, right? And it's really simple to build. So, if I go into Opera, you can see in my community, I have some updates here and there that I've been updating my community how the bot is doing. But it was really simple to build because all I had to do is go to uh this platform called alpaca.markets and then you just sign up. Okay, so I just signed up. This is a paper trading account. And you see, let's say it has 100K. And if I scroll down to the bottom, there's this thing that says generate new keys. So, I'm going to hit that button. It's going to generate new keys. I'm going to accept all the stuff. And you see it gives you an endpoint for the keys and a secret. So, what you want to do is right here. So, let's say you go to your Discord. I like using Discord for this uh use case because what you can do now is you can make a new text channel. My thing isn't working. Uh trading and then right now I'm going to give it all that stuff. So this is alpaca. Let's give it the end point Discord key and Discord secret. Okay. So, all right.

Now, we're going to give it a strategy. So, uh, let's say I won't mind owning Tesla for let's say 10 years, right? So, I'm going to use a strategy and tell you guys about it. Um, so what we're going to be doing is trading stuff like Tesla in the wheel strategy and simply explained the wheel strategy that I want to implement is the way to generate income from a stock that I like and won't mind owning forever. Right? So how it works is basically step one, we want to sell a cash secured put. So let's say Tesla is trading at 250. You say I'd be happy to buy Tesla at 230. So, you put an option at that 230 strike price and someone pays you, let's say, 300 uh premium. That means they're paying me just for making that promise. Now, two things can happen. Tesla can stay above 230 and they put an expires worthless and we get to keep that 300 they paid us. That's easy money. And if Tesla drops below 230, then we would get assigned those shares, meaning we're forced to buy 100 shares of Tesla at 230. But since we like that shares, right? Uh we already collected that extra 300. Okay. Step two is we sell a covered call. So now that we own let's say 100 shares of Tesla at 230. Tesla is sitting at let's say 235. You say I'm fine selling my shares at 260. So you sell a covered call at 260. And someone would pay you another $200 in premium for that. Okay. Again, two things. Let's say Tesla stays below 260. that means the call expires worthless and we get to keep that extra 200 still holding our shares. And if Tesla does blast past 260, that means our share gets called away, sold at 260. And we just made money on the shares and kept the premium. And then step three is we just repeat. So whether this our shares get called or not, we just spin this wheel again and sell another put. We collect more premium, rinse and repeat. The big idea is we're not trying to predict where Tesla goes. We personally I personally like the stock. So, I'm just collecting rent on the stock that I already like. And this works best with high quality stocks with decent premium options. And Tesla works really well for this. All right. The risk is if Tesla absolutely craters and it goes down like drops to 150, we're sitting on a massive loss. That's why you only run this on stocks you generally wouldn't mind holding longterm. So, we sell puts to get in, we sell calls to get out, collect the premium the whole way, and that's the wheel strategy. All right? So, that's how I want you to be trading. If there's other stocks like Micro Strategy, buy those too.

Cool. That was me dumping a lot of knowledge from stocks that I've been doing. Um, maybe Okay, it's not accepting all the my big rant. So, what I'm going to do is if I just paste it, uh it should be fine. Nope. Okay. Okay. Okay. So, I'm just going to make this go and then just send this as another message. And um OpenClaw has a really good uh queuing system. That means it's going to hopefully pick this up and now it's going to use these keys for me and start trading this wheel strategy as we go on. All right. And if you guys want to follow along, join my community. I'll post updates on what's going on. But this is a really cool strategy and how we can do automated trading with the bots doing all the menial tracking this stuff on our behalf right and it's really cool and you can see right now we gave programmatic access to this account and in the same way I've been doing this in my community and we see you know my bot has been buying and selling buying and selling for a couple days where I've been updating people on what's going on and this is actual 10,000 I put in not a money trading account and that was pretty fun to do. Let's see this. Uh explain back. Okay, cool. Whatever. Uh, if you want, yeah, set up the trades. And now I'm going to be like monitor the market every day. Check every six hours or so and make sure you do the right thing um every day. Okay, so that's your crown job. And check the market. Great. So um now I set the cron job and market and I am should be good to go. And now I should have an automated trading bot with openclaw that lives on my Mac mini has access to an account and is doing all this stuff. Great. Okay. So hopefully you guys learned how to build an automated trading account. Don't take this as financial advice. This is just just for fun and I'm showing you how to do the stuff not what to do. Uh, do with this information what you will, but I thought this was pretty fun to say. Oh, I see it's already setting up its own cron jobs and monitoring the stuff. Cool. So, let's see how this uh works over the long term.

Okay, so for build seven, I'm going to show you how I put open claw on my metaray bands. And this is going to be opening up so many doors uh to just future businesses that we can't even think of right now. and I'll show you a quick demo and then get into why this is so impactful. All right, and then we'll set this up. >> My car has been struggling to accelerate. What's going on? >> From what I see, your air filter is clogged. That restricts air flow to the engine and kills your acceleration. What car is it? >> This is a 2021 Honda. >> Got it. I found a replacement on Amazon for $16. Want me to order it? Uh yeah, add it to my cart. >> Sure. Adding that to your Amazon cart now. >> All right. So, I just asked to do that. Now, this right here is my browser. And you can see right here, it's actually controlling my computer to go find the Honda 2021 air filter or whatever. And it should put it in my cart. And you can see it's actually been sending me live updates and pictures of what it was doing. And it's saying the $15 right here. So if I go, this is like my Mac Mini. Uh I can see that, you know, it added this to my cart right here. And you see the cool part is that I didn't know what was wrong. The AI saw what was going on. It diagnosed it. It ordered the fix all through my glasses. And I'll show you a few more examples later in this video, but the Amazon part might seem simple. Maybe you can set up an automation to order something here and there. But the really cool thing that happened is that I didn't have to tell it to go to Amazon. I didn't build an order from Amazon workflow. The AI had full access to that computer and it decided that Amazon was the right move. It could have just as easily texted a mechanic, searched a parts form, or pulled up Honda's service manual, but it chose the best action based on what it saw and what I had asked it. And that's what makes this agentic. The AI decides what to do. You just give it the permission to act.

So, how did I build this? Well, I took the meta ray bands, the ones with the camera built in, and I put this autonomous AI agent called OpenClaw on them. And OpenClaw has been blowing up recently. And for good reason. It's an open- source AI agent that runs locally on your machine. And it can see your screen. It can control your computer. It remembers every conversation you guys have had and takes actions across the browser and your apps. And what I did was I bridged OpenClaw with my glasses. Now, Open Claw just doesn't live on

my computer anymore. It literally lives on my face. And when I have it on, the AI doesn't wait for instruction. It proactively can act. It has its own reasoning. It decides what to do based on its context, its memory, and what it sees happening. And the same AI on my glasses is the same one I've been texting on my phone and the same one that's also on my computer. It's like a partner that's always paying attention.

So, in this video, here's what we're covering. What these glasses can actually do, like the full range of capabilities with some examples, where this technology is headed, and how you can actually take advantage of it. And finally, how you can set the whole thing up for yourself.

But why does any of this really matter? You know that scene in the Matrix where Neo plugs a cable into the back of his head, wakes up a while later and says, "I know kung fu." Physical work is about to get something similar, not through brain implants, but through real-time AI guidance. Much of the AI conversation focuses on which desk jobs will get replaced. But for physical work, stuff like field services, manufacturing, and healthcare, AI can't yet act in the world. What it can do is see, reason, and guide the human who does. Imagine wearing a small camera while an AI sees what you see and talks you through the job. "Turn off that valve." "Use the half-inch wrench." "That part looks worn. Replace it." And so on. Instead of needing months or years of training, workers can be effective immediately with AI coaching them and accessing new skills when needed.

And that's the team at Y Combinator. And they're literally investing millions into this tech. And what they're describing is exactly what these glasses kind of do. Accessing new skills when you need them and on the spot, which means you're not limited to what you already know. You have every skill, every language, every piece of knowledge available to you without having to ask.

And the reason this is possible right now is because two brand new technologies are converging. So recently, Google actually released these multimodal AIs. That means AI that can see, it can hear, and it can process everything that's being thrown at it all at once. And when you combine that with something like OpenClaw, which gives the AI the memory, the computer control, and the ability to take actions, you get something the world's never seen before.

This is what I would call actually smart glasses. And I mean that as an acronym because each letter represents a real capability. So first, they can see. The AI sees what you see. Like, you can point at something, you can say, "What is this?" It identifies the object. It reads the label. It goes in, does some research, and helps you figure things out on the fly.

And M stands for memorize. The AI stores and recalls information. So let's say you store a business card on Tuesday and you ask about it a couple weeks later. The AI is going to remember. And it's powered by OpenClaw's persistent memory, which means every interaction gets logged and indexed into a structured memory that the AI can search and reference at any time. And the cool part is that it's stored locally on your computer. So nobody else has access to it. Actually, let me show you.

"All right, so I just started this and then watch this. So, hey, uh, what was my, um, biggest task for today?"

"Understood. Your biggest task for today is to film the VisionClaw video."

Okay, cool. I mean, it's something small, but this is I just wanted to show you how it can remember things about what I've told it from the past.

Then it can also act. OpenClaw literally has access to its own computer, so it can click, type, navigate apps just like you would. It also reads any language, which is kind of cool because let's say you're abroad looking at a Japanese menu. The AI can see the menu, can translate it for you, look up what's the best-rated stuff here, and give you recommendations.

And the most applicable part is the glasses can literally guide you step by step. If you're looking at, like, let's say a circuit board and you have no idea what you're doing, like, "How do I replace this capacitor?" It walks you through it, watches your hands, correcting you in real-time. It's all of the above working together. The vision to see your progress, the memory to see what step you're on, and the language to explain it clearly. And it can even text you diagrams and reference material if needed.

So, let me show you what this looks like when you actually stack these features together. And let's go through some examples of what it can really enable. So let's say you're walking into a meeting and you say, "What do I know about this client?" The AI can pull up your last three conversations with them, their company details, opens up LinkedIn on your laptop, all before you sit down.

Another example is, let's say you're inspecting equipment on a job site. The AI can see issues, identify the parts, the numbers, logs the problems in your maintenance system, and generates a report. You just have to look at it and train it beforehand. The paperwork would do itself.

Or let's say you're in a store and you see a product you want. You can say, "Hey, order this." And it'll read the label. It'll find it on Amazon and order it. Your hands never have to do anything.

But the thing is, where we're at has its cost and limitations. Now, I want to be straight with you. This technology is really cool, but it's still new. It's experimental. Like, just for me to get this demo working took a lot of attempts. It's cool now, but I had to work a lot on it. But the thing is, this is the worst this tech is ever going to be, and we're just going to be going up and up from here.

Also, only a handful of people in the world have access to this working, and it's not even free to run, right? So, actually, let's talk about the costs because the glasses themselves, the Meta Ray-Bands, run about $300. That part's straightforward. But the AI models, when we use it, that's where it gets interesting. So, when you're streaming live video or something to like Google Gemini, you're paying per second of processing. And depending on the model and how much you use it, that can add up fast. We're talking potentially hundreds of dollars a month if you're running it all day. And then the battery for the glasses is another concern. And if you're running OpenClaw with raw API tokens, let's say Claw, GPT, or whatever, those costs can also spiral. And people have burned through thousands of dollars just leaving an agent running.

So I actually found a hack to help alleviate parts of that. And actually, the founder of OpenClaw recommends this too. And instead of paying for API tokens directly, you should probably just use a ChatGPT Plus subscription. And OpenClaw can route those tokens through that subscription instead of eating the raw API credits. So instead of paying per token, which gets really expensive, and you're paying for, let's say, a flat $20 a month for ChatGPT and OpenClaw, and right on top of that.

The other limitation is that this requires like a Mac, an iPhone, and some technical setup. This will get easier over time. The tech will get cheaper, but right now this is where we're at. And for the people that are willing to put in a little bit of time, you're going to get something that nobody else has access to.

Now, if you want help setting this up, I actually started a community for this exact reason. The people in this community are the ones who actually inspired me to do half the stuff you saw in this video. We've been having calls, testing these use cases, sharing setups, and they're building the stuff alongside me.

"I will say I was so happy I found you."

"I literally spent the day before like 12 hours trying to get it set up. If I had yours the day before, I wouldn't have spent the 12 hours because I kept getting hung up on some of the simple things."

"So, I would say to you, I know everyone's like high-tech and and I'm I'm about making the money with everybody like what they're saying, but I think the a lot of people are leaving the beginners."

"And so, how you broke it down, that's why I kind of joined you was simply because you made it so simple for the everyday person to be able to understand it and how to be able to function and use it. So, that's why I'm where I'm at literally is simply because of what you've what your video."

So, if you want to set this up with people who are actually doing it, the link's in the description below. All right. So, let me help you set this up. So, before we get into the actual build, let me show you how the system kind of works at a high level. And there are three components. One, which is the glasses, the Meta Ray-Bands, the built-in camera and microphone. They're the ones that capture everything you see, but don't do any of the AI processing. They're just the sensors. And the bridge is your iPhone, and it's going to be done through the Meta View app, and your phone's going to receive whatever the camera is seeing and stream it into Google and OpenClaw.

Then, Google Gemini, you're going to need a key, and that's the brain. That's where everything kind of happens. And the camera feed from your glasses is basically piped into Gemini's multimodal API, which is what gives AI the ability to see, understand what's in front of you. Then we're going to need OpenClaw to be set up on something like a Mac Mini or a VPS. This is what gives AI the memory, the computer control, and the ability to actually take actions. And when Gemini processes what you're looking at, it's going to shoot that over straight to OpenClaw, and OpenClaw is going to execute it on your computer. Stuff like ordering stuff, sending a message, pulling up information, whatever the task may require.

All right, so let's build this. I've also got a setup sheet linked in the description below if you want to follow along step by step. I'm going to be walking through this live, but this sheet has every command and link you can follow along with.

All right, so to help you set up VisionClaw, these are the steps we're going to take. So let's dive into step number one. Okay, so what we want to do is first go into our Meta app. And then from the Meta app, we want to go into our settings that's located on the bottom left. And after we go to settings, we want to scroll down and scroll down until we find, there we go, App Info. App Info. You want to make sure you have developer mode turned on. And to do that, what you need to do is just tap App Version five times. And then after you do that, you should be good to go.

Great. So the next step is we just need to install Xcode. But the nice part is that we don't have to do anything because we'll just get OpenClaw to do the rest. So, I'm going to open up my Spotlight search again. Then I'm going to go to the App Store and then I'm going to search. And in search, I'm going to do Xcode and then I'm going to download this. All right. Okay.

So, the next step is let's go get our Google Gemini API key. And to do that, right here I'm going to go to ai.dev. And right here on the bottom right here, I'm going to hit "Get API Key." And then I'm going to hit "Create API Key." And then let's call it, um, OpenClaw and hit "Create Key." And then all we have to do, you see we just created it, OpenClaw. And I'm going to copy this API key. And that's it. And I'm going to show what to do.

What we do next is in Google, what we're going to do is type in VisionClaw. And VisionClaw, the first GitHub repo. You're going to click this. And all of this you just saw and what I did was made possible because of Shalu. He made this VisionClaw repo and I just did a lot of development on top of it. But what we want to do first is now we want to come here, take this, copy this. And now everything we do is just talk to OpenClaw. So I'm going to paste this here. And then I'm going to just speak to it.

"All right, hey, so can you help me set this VisionClaw repo up? I have Xcode downloaded on my computer. I want you to set that up and then also I am going to give you my Gemini API key. I'm going to paste it here. Can you do all the linking and everything for me? So after that, let me know what's next so I can just connect my iPhone and go."

So after that message, if you open up Xcode, what you should see is that it already built out a project for you. All right. So I'm going to open that up. And it has a bunch of code that I don't need to know what it is. But what I do need to do is now I need to take a USB-C cable. So I'm plugging this in. And then I'm plugging this into my iPhone. Is now what you want to do is you want to go here. You want to hit this and then you're going to look for your iPhone. So I'm going to select Simon's iPhone. Okay. And then you're going to hit play. You might have to just like put in some permissions and whatnot. But what's going to be happening right now and then you'll see. And then you hit play. Okay. And then I already have the app downloaded, but I'm still going to replace anyway, just so you guys can see. And what it's doing, it's downloading this app onto my phone. So, right here, if I hit this, and then you can see this app will be here, which means now I can start my iPhone. And this app is the same thing as we just experienced. So you can see.

"Hello."

"Hello. How can I help you?"

See, that's cool. Uh, and you can hear it happening from my glasses as well, cuz I'm wearing it. And that's it. As simple as that.

Okay. Now that we learned how to build so many different things, let me show you how we can actually take these builds and actually make money from it and implement it in business. And to start off, let me actually show you this clip.

"I've been through every single technology, you know, event and evolution, and this blows them all away."

Now, how you implement it in business is a whole different issue. Like literally, when I was 24, I was walking into companies who had never seen a PC before in their lives and explaining to them the value and having these guys going, "Well, son, I got this receptionist right there. I got that secretary. I'm never going to need that ever." Right. And but then, you know, my business then was helping them figure out how to implement it to give them an advantage.

"There are going to be integrators that particularly young kids like when I'm telling my kids who are 15, 18, 21 or 19, 21 and kids going into school, what should I do? What should I do? I'd be like..."

"AI research."

"Learn all you can about AI, but learn more on how to implement them in companies. Right. Because to your point, companies don't understand how to implement all that right now to get a competitive advantage. You got the head of Microsoft saying software is dead because everything's going to be customized to your unique utilization, right? Or usage. Who's going to do it for them? Particularly small to medium-sized businesses. There are 33 million companies in this country. 30 million of them are solopreneurs, right? Um, single-person enterprises. There are only, you know, there are millions of companies that have one, five, 10, 50, 100, 500 people that aren't going to have AI budgets, aren't going to have AI experts. This is where kids getting hired coming out of college are really going to have a unique opportunity. If you're spending your senior year in college right now, your senior year in high school, even whatever it is, your excess time, and you're learning the difference between Sora and VEO, and you're learning how to do all this video, you're learning how to customize a model so that then you can then walk into a company and say, I understand your business as a shoe company selling shoes at a retail store or selling and selling shoes online. Let me show you how to benefit you. That is every single job that's going to be available for kids coming out of school because every single company needs that."

You see, that's Mark Cuban, the guy from Shark Tank. And you can see that the potential is obvious, but the path to getting there isn't clear for most people. That gap between what technology can do and what regular people can make it do is where the money really lives. Like, for example, WordPress, Shopify, all these things are free, but people still pay thousands for someone else to set it up for them. Why? It's because they don't want to learn it themselves. They just want the result. The same thing is happening right now with AI agents and OpenClaw. And OpenClaw is literally open source. It's free. But 95% of the people who could benefit from an autonomous AI agent will never be the ones that set it up for themselves. They don't even know what a terminal is. They don't they don't know what an API key is. They have zero interest in learning. You are now in that 5% that can give them that result.

So with that being said, what can you sell? Well, there are four easy business models which you can take and sell from what you've just learned. Each of them have different margins, different levels of effort and scalability.

And Model One starts with just done-for-you builds. So you'd build a custom agent or system for a specific client. So, you'd have to do the discovery call, scope out what they need, build and deploy it, and then hand it off to them. This is probably the highest price per engagement. I've seen a lot of people sell this for $2,000 to $10,000 depending on the company. This is the most work per client, but it's the right model to start. But if you're just starting out, it's the right model to start because with every build, it's going to teach you something new about what client needs are. You're going to get paid to learn. And most of my community members actually get started here because you can just do this once, and then every subsequent build gets much easier.

Okay, Model Two is pre-configured packages. So instead of building it from scratch every time, you can create a template agent and set it up for a specific niche and sell it to multiple clients. So think the content creator agent package or the real estate agent package. The same core configurations but just a slight bit of personalization for each client's voice and their workflows. That would be kind of like a mid-tier. So around $500 to $3K because you're not really starting from zero each time and your operation cost is a lot lower where you build once and then deploy for others. The first build maybe takes a bit of more time, but the second and the third gets a lot easier.

Okay, Model Three is productized services. So standardized offer with fixed scope and a fixed price. So let's say for $1,500 a month, your agent monitors your community, it drafts content and handles your email triage. This is the easiest business to scale. But you really have to do research and you have to know the exact workflow that gives the value for the person that's buying from you.

Okay, Model Four is a SaaS-style product. So, for example, this is actually a Claw wrapper. It's built on top of platforms like OpenClaw, but for specific verticals. There's already hundreds of OpenClaw wrappers in the market. You can see some of them right here. But this model requires more upfront investment from your time. It has the highest scalability, but it's also the most fragile and easiest to break because AI moves so fast. People start should start with either Model One or Two and work towards this later after they really understand the nuts and bolts.

But after that, what I want to really, really emphasize is when you start out first, pick one niche. Like the biggest mistake is trying to sell to everyone. Like, "I build AI agents." That means nothing to most business owners. They don't even know what they are. It's vague. It sounds expensive. But if you can say something like, "I'll build you an AI system for e-commerce stores that creates 60-plus ad creatives per week." That's an offer somebody will pay for. It's specific. It has a number in mind and it describes the result they actually care about. And niches that are working right now that I've been doing a lot is coaches, e-commerce stores, real estate agents, marketing agencies, content creators, and just local service businesses. The best agent business won't come from the best engineers. They'll come from people who understand their target customer's daily workflow better than the customer understands it themselves. What that means is you don't need to be the most technical person. You just need to be the person who understands what a fitness coach struggles with at 6:00 a.m. and which parts can maybe be automated.

Now that you understand the ways we can sell these, what are different pricing models we can leverage? Because different clients and different industries like different pricing. So the easiest one to start up with is a setup plus retainer. This is the most common and most profitable structure, which is basically just a one-time fee plus a monthly retainer. Some setup fees can range from $500 to $1,000 and multi-workflow systems when it's actually integrated and working day-to-day. You can take that up a notch each time you build something new. And retainers can range between $200 to $1,000 per month depending on the value you're actually getting for your client. And what the retainer is there and how you pitch it is because you're saying, "I'm going to be monitoring this. I'll be updating it as it gets better, adding new skills, troubleshooting, optimizing it." And the retainer is where really the money is. It's recovering revenue that compounds as you add clients. For context, AI agencies across the board are charging around $500 to $5,000 per month for ongoing management.

And the easiest one if you're just starting out is value-based pricing where you don't have to price based on your time. You just price based on what your clients get. So I'll give you some math. A content creator, let's say 15 hours per week on content production. Their value and then they value let's say their time at $100 per hour. That's $6,000 a month in production cost. So if you set up an agent that cuts their production time to 3 hours per week, that's $4,800 a month on savings for them. You don't have to charge that much, but you can charge let's say $1,000 at $500 a month for maintenance. It's a fraction of what they save and the ROI kind of sells itself. The framing is always, "Here's what this costs you today, and here's what this costs with the agent." That difference is basically yours.

The monthly retainer deliverables is basically the monitoring and uptime checks, the troubleshooting when things break, adding new skills, optimizing workflows based on any output quality and any updates to OpenClaw or any connected tools when they have new versions. If you write this into your proposals and send it as a one-page document, the client can exactly see why they're paying you and what they get and makes it a lot easier to mull over the price.

Now, if you want to find clients, strategy number one is just use the thing to sell the thing. Where this is a meta-play and you actually, we just built a content creation agent. You can make content about that, have a link to your calendar, and it's very easy to get content and leads from there. You can also build, record yourself building the morning briefing agent, show an output for a specific industry, posted on LinkedIn, Twitter, YouTube. And these screen recordings of your agent doing real work are the best marketing material that'll ever exist for your service business because you're not telling people it works. You're literally showing them what you're going to do for them. People may watch it, may say, "I need that." And they're your leads. They come to you. You don't pitch them, and they see your product working. This is the most effective lead generation strategy I've seen in my community that works, and the content demonstrates the product. So you don't need a pitch deck or anything.

So strategy number two is cold outreach with proof. Now, if you let's say you pick your niche, you can find 50 people in that niche and send them a personalized message. So maybe this is on LinkedIn using Sales Navigator or just search for them. Like, don't tell them something like, "Hey, I build AI agents that nobody cares about." Say something like, "Hey, I noticed you post content five times a week across LinkedIn and Twitter. I build a system that researches topics, it drafts posts in your voice and creates a carousel on all on a daily schedule. Here's a 30-second video of it running. Want me to show you how this works?" Very specific, relevant to their actual work, and it includes proof. And you can take the old builds we did and then you sell that. That's the recursive part. Your agent can actually help you with this too. Have your agent do the researching for the prospects. Personalize the outreach messages and track who responded. Use the tool to sell the tool. One approach that's working really, really well is when you pick a vertical and build one great demo for that vertical and then targeted outreach to 50 prospects in that space. Even a 5% response rate gets you conversations and some sales.

Okay, strategy three is you go find a niche community. So every industry has communities like Facebook groups, school groups, Discord groups, subreddits, Slack groups. If you want to join those target niches and you answer questions and share what you're building, if you actually genuinely help people and you don't have to pitch, you just demonstrate. And when someone says, "I spent 4 hours a day on managing my email or building this carousel," you could say like, "Hey, I built an agent that that handles that. Here's how it works." You post a screenshot or a video and you share the workflow, and the client would naturally come to you because you've already shown them that you understand the workflow and you're part of their sector.

But the biggest thing I want to make sure you understand is here's a specific play for your first client where you find someone whose work you actually understand really well. Offer to build them a prototype for free or at a steep discount. Tell them, "Hey, I want to build this for your workflow. Help me test it and give me feedback so I can make it better." Okay. After you have that done, they get a useful system at a low cost, and you get a study, a testimonial, and deep knowledge of their workflow. Then what you can do is you can charge the next person full price, and you ask your number one client, the first one, for a referral.

So just to walk through a concrete example. So let's say your target customer creates content regularly three or more times a week and spends 10 hours a week on content production. He has an established voice and style that they want to maintain, and they use multiple platforms. So X, LinkedIn, maybe a newsletter, maybe YouTube. You're not selling to people who want to start who want to start creating content. You're selling to people who are already doing something and drowning in work or maybe want a bit more freedom. And the pitch you do is you go from spending 80% of your time on production to 80% of your time on ideas.

Most people will actually learn about autonomous AI agents like OpenClaw in 12 to 18 months. Now you have everything you need to learn it today and actually start making money from it. Most people will spend months trying to figure out the technology. You already built seven systems by now. And with that head start, that matters. Not because the knowledge is secret. It's all open source, but because the experience compounds. Every build you do makes the next one much, much, much faster. And every client teaches you something.

"All right, Andrew. Cool. So, I've been hearing you were selling some OpenClaw offers. Can you take us through, you know, what your offer was?"

"Yeah. Um, so first thing, the first thing I did was because I didn't know if I knew what I was doing. Like, I knew what I was doing for me, but I didn't know what I was doing for other people. Um, and so I had a couple of friends that are business people that I was like, 'Hey, look, this is what I've done.' Um, they were able to..."

"What kind of business?"

"Uh, so one was in commercial real estate, um, and the other one is in roofing. Um, and they're in two, one's in Michigan, one's in Alabama. And so I was kind of showing them what I was like my call my Stfan. and I was like showing them all the things that I could do. So anyway, they, um, I kind of used them as guinea pigs to kind of do it and I figured out what to how to systematically do it and, um, I use Tactic. I don't know if people I highly recommend that everyone has something that can record the transcript because then I could take all of those different calls that I had with my friends, how I was doing it, and then I put it into Claude and I said, 'Hey, take, help me systemize this process.'"

"What was it that they got value out of like at the end with OpenClaw that..."

"They, so essentially, so they just have an AI assistant now, right? That's like, that's that's connected to their their Gmail and things of that. And so that they can, um, you know, from their phone, they can say, 'Hey, you know, they've built, I've helped them build out daily briefs because I really like the fact that it can while I'm asleep or I'm away, um, it's kind of, you know, because a good executive assistant is like the front line or the first line of defense, right? When it comes to your schedule, that it knows what's going on, what's coming up, as well as, um, by the way, here's who has emailed you over the night or, you know, you were in a meeting and this person or that person called, gives you an urgency, um, and you start to, um, really own your time and own your schedule because I think a lot of people, they, um, they base their their time and schedule off of what the most urgent thing is. And so there's not a lot of control on their part. Does that make sense?"

"So it's kind of like an assistant, executive assistant, the morning daily briefs, some follow-ups, those kind of things, right?"

"Yeah, that's the main thing we're using for right now."

"What did, what were you able to like charge them for this?"

"Uh, so then, um, one paid me $500 and one paid me $1,000. But now I've done three more and those have ranged between $2,000 and $5,000. Um, but just to be transparent, honestly, it's not like you're going to sit down and I, I, you know, I keep trying to better systemize it so that it's like people can get, um, where they want to go because at the same time too, like now, um, obviously I can use Stfan to, um, research things while I'm away and, um, and to, you know, I give it tasks that it will remember to do, um, which is really, really, really helpful."

"What kind of..."

"Uh, like if I'm going to be talking with a certain company, um, I ask for it to go and they dare scrape that company or the people that are going to be on the call and..."

"You're using the tool to sell the tool."

"What is it?"

"Yeah. Yeah. Exactly. And I'm like, 'This is what...' Yeah, exactly. That's a good, that's a great point. I didn't really put that together, but yeah, you show them like, 'This is what it is and this is how I do it.' Um, and even, but even people beyond, you know, buying the the service and whatnot, it's just such a great thing to have as far as really being prepared. I think that that's the one thing, um, is that I love that I now feel like I own my schedule more. Um, and that I feel more prepared in all of my meetings. Um..."

"That's really cool."

"That's I think maybe a lot of people struggle with this part as well. So now let's say you learn how to build these kind of things, these cool assistants, and there's a lot of use. And the way you were getting these sales is just demoing it. How were you able to find these prospects or leads in the first place?"

"I always say, listen, this is age-old thing of sales. Use your warm market. And the cool thing is is that there's always going to be people that are willing to, um, help you figure out what you know while you're you're figuring it out, right? I wouldn't recommend charging those people. I mean, unless you can..."

"I guess you should not charge, right?"

"Yeah, I would I would highly recommend that just because, you know, you're going to, it's not an if, but a win that you're going to hit some kind of a wall that you're going to have to figure out, and you'll always be able to figure it out if you, especially if you use the right, you know, AI AI tools. I use ClawChat, um, and and I even created a project specifically for, um, this bit this kind of business so that all of my, um, interactions with all the people, you know, my friends and and the clients because then it of course remembers everything, right? And so then I have the ability when I'm working with someone new that if I'm like, 'Oh wait, I know that we've run into this specific issue within that,' um, project, it will be able to pull that out and be like, 'Oh...'"

"No pressure. You're helping people, you're getting onto conversations and then seeing where that goes. That's like a good..."

"Strategy. Exactly. And then just constantly asking Claude, um, afterwards to update the system that it's created because I, you know, it's like it keeps kind of creating a playbook for me to to use, um, so that, um, I don't particularly send that playbook out to people because you're kind of giving your sauce away and I would recommend that. Um, but you can definitely do like a, um, a like a quick, you know, a lot of times when you buy a TV or something like that and it comes with a quick manual where it has like, you know, here's some tips and some quick like ways just, you know, to cut to this or to do that. Um, um, I asked Claude to create that for my new my new clients that I that I send to them so that because a lot of times they forget just the smallest thing. 'How do I do that? How do I restart the gateway?' or things of those things."

"I got you. Yeah. So..."

"So what's next? So, now you're selling these OpenClaw agents. I remember you were also selling voice agents before. You're kind of expanding your pool in that way, right?"

"Uh, yeah. And so now we want it to be because, you know, you and I both know that where everything is leading or heading to is that we are, you know, you have AI assistants, and then you're going to have AI colleagues, and then essentially you're going to have just AI worker bees that report back in a very much, you know, you, you know, you've heard me use the metaphor of like, I'm the conductor, they're in the symphony, they're in the orchestra, and they all are playing their own individual parts. They're doing the thing, and they're just reporting back to me. Um, and and essentially I'm trying to take myself as much as I can out of the..."

"Loop. Agent workforce. That's what you're trying to..."

"Exactly. Right. Because everyone has to understand that's where we're heading. You know, I'm sure you've heard this, and and I'll just repeat it for those who haven't. Don't build for now, right? Because if you start to, if you take what is now and you're trying to just build the AI for the for for now, um, by the time you kind of get it really going, you're going to be behind. So build, think like where are we going and what's going to look like in six months? Build for that. Build for six months from now in that in that way because then you'll start to realize that that's what puts you ahead because you know what is considered ahead becomes right in the moment, really quick now, you know what I mean? Like very few people can say I'm ahead for very long. If you don't believe me, look at how, you know, Gemini will put something out or Anthropic will put out a new model, and all of a sudden, Chat, you know, GPT-5, 4 comes out, and all these updates to Codex, like no one really is, you know, is quote ahead in the AI game. But the best thing to do is to is to really kind of look at where we're going. And where we are going is the agentic workforce that we as humans work with."

"Awesome. That's really cool, Andrew. Thanks for sharing that. Um, you know, hopefully these guys are a bit more inspired. So, it's not just like there's so many people taking advantage of like OpenClaw, ClawCode, all these just to make life a lot easier and also make some money out of it. And I love that."

"Yeah. I just want to end with this. There's it's that's so great. Anybody that if you really want to do it, first of all, you got to put the reps in. AI makes it easier and it makes it more exciting. But to think that you're going to be able to sit down and an hour later you're going to just have all this, you know, amassed all this, you know, incredible knowledge and and use cases is is is wrong. Um, but also just learn where you are effective because I think a lot of times people kind of do it backwards where they try to sell the excitement of the AI and like, 'How cool is this? And how cool is this?' Um, and then at the end of the day, they don't make a sale. And it's because you're selling cool, but you're not selling outcome. And that's all that people care about. As far like, if I want to see a cool trick, that's cool. But if I want, if I'm a business, I'm going to write you a check. I don't care how cool the trick is, right? Show me the outcome. Show me that it's going to be better than what I could do by myself or somewhere else. Um, and you're going to get, not only are you going to get people to write that check quicker, but you're also going to get, um, uh, people to, um, give you referrals. And that's really big."

"That's awesome, Andrew. Thanks. Yeah, this is really good advice. You guys heard it here first."

"Yeah."

So, what to do next, right? So what I want you to do is pick one build that matches what you already work on. Run it for a week. Get comfortable with it. Then pick one niche from this section. Maybe a type of person whose workflows you understand. Identify which of these seven builds may help them out. Then build it for yourself. Use the tool to sell the tool. And then you can join the community. The link's in the description. That's where you'll get help. You'll see what others are building and actually get access to new builds before anyone else. And the biggest thing I want to communicate is that the way people are working now is changing completely. Every company will need autonomous AI agents like this. Most of them have no idea how to even build them. And the people who can are going to be in huge demand for a long time. Now you know how to build these things, how to debug them, how to sell them. I want you to go get started. And if you need any help, I'll be in the community linked below.