What Is DeepAgent (Abacus AI)? 🧠 Full Tutorial & App Building Example

The fact that I was able to build this entire app all in one go, that's mind-blowing.

Like, this is saving me so much time already. Oh, interesting. So, I asked it to debug and right now it just created a virtual version of itself and it's autonomously actioning the computer for me, fixing it. It's kind of surreal watching it go through and click. Oh, there we go. See, it does work.

Deep agent is one of the best agentic AIs I've ever used. You can send an agent to go work on tasks without any assistance, including building an entire app from start to finish and deploying that app directly within Deep Agent. That means you can build an app, debug it, and then share it with others to get feedback on your own custom subdomain through Abacus AI. App building, website design, deep research, document creation, automation workflows. There are so many things you can do with Deep Agent. All you need is an idea and Deep Agent will help you turn it into a reality.

Hi, my name is Callum, also known as Wanderloots, and welcome to today's video on Deep Agent by Abacus AI. I've tested a lot of Agentic AI over the last 6 months as I've been building various apps, and I have to say that I was very impressed with the thoroughess of Deep Agent. Most AI tools, the Agentic Coding Systems, Vibe Coding, can build a front end, no problem. But Deep Agent was actually able to build the front end and connect it to a back end with full database, authentication, and payment integration. Of all the tools that I've tried so far, none of them have been able to do that complete full stack app build in one go. Of course, the app wasn't perfect on the first go, as you'll see shortly. But the true power of Aentic AI is not with its ability to code the initial system, but its ability to debug and fix problems and solve errors so that you can go through and build a fully functioning application.

In today's video, I walk through a brief overview of Deep Agent and how Agentic AI differs from chatbots like Chat LLM. I also share a series of practical examples of how to use Deep Agent. Then I get into building a fully functioning chatbot application from start to finish. I also deploy the app, give some tips on debugging, and give you a demonstration of Deep Agent Desktop, an editor that gives you more advanced control over the coding process.

If you find this video helpful, please like hype and subscribe as your support really means a lot to me. Also, if you want to try out Deep Agent, please use the link in the description below, deep agent.abacus.ai/cwl, AI/CWL as using that link helps me out a lot. Now, let's dive into Agentic AI use with deep agent.

So, what is deep agent? So, deep agent is part of Abacus AI which also includes chat LLM that I recently made a full video about. So, if you're interested in learning more about the chat LLM side of things, I recommend checking out that video. And effectively, deep agent gives an agentic element to Chat LLM. So, what does that mean? Well, if you think about it, most of the AI that we think about these days, like ChatGpt, Gemini, Claude, they're all like chat bots. And that's a very passive system. You ask a question, you type in a prompt, it gives you an answer, and you can send it to go do tasks like search the internet, but it doesn't necessarily daisy chain and connect those tasks together to do things autonomously. That's where Agentic AI comes in. Aentic AI is an active system. It lets you build apps, generate documents, and it can have a series of complex tasks that all connect together and can operate autonomously.

So, I don't want to get too deep into Aentic AI today. I'd rather just show you how it works. But, if you are interested in learning more about Agentic AI, please let me know in the comments and I'm happy to make a video that explains more deeply how Agentic AI works and what it is.

So, as part of your $10 a month subscription to Abacus AI, you get access to Deep Agent Chat LLM and Deep Agent Desktop, which I'll explain in a little bit. And the key here is that you're not just getting access to one state-of-the-art model. Instead, you're getting access to all of the state-of-the-art models. In chat LLM, for example, there's route LLM and it will automatically go through and select the best model for your particular use case so that you can route between the different AIS, the different LLMs and solve your problem as effectively as possible. As part of deep agent, you also get access to all those models. And the benefit here is that different models have different specialties. They have different levels of effectiveness for different types of tasks. So, by having all of the models in one place, the Deep Agent can switch between the different models depending on what you're asking it to do. It's useful for anyone that's looking to be more productive, effective, and efficient with their time without having to spend money on a whole bunch of different AI tools. And there's a lot of different use cases that you can get into here. You can check them out at deep agent.abacus.ai. They've got a full list of all the various examples along with prompts that they use to create them. So, I recommend checking this out. And I'll show a few of these as we go. But now that you have a good idea on what Deep Agent is and how it works, let's get started.

So, when we get into the homepage here, you can see that it's got the classic chat input here. So, this is where you can task the Deep Agent to go do things for you. There's an FAQ button here if you want to click on how to. And there's the option to connect different tools, including MCP servers. So basically what this means is you're able to connect for example Google Drive and bring in files to Deep Agent so that you could take the papers or articles that you've written or images that you've stored there and you're able to give them to the agent so that it has context on what it is you're trying to do. You can connect your calendar or your Gmail and it would have context on your meetings and your correspondence so that you could for example have it automatically summarize the emails you received last week and prepare a task list for yourself next week. And all of this could happen automatically because we're dealing with agentic taskbased AI.

Then we also have the option to configure MCP. So MCP stands for model context protocol which basically is a standardized way of sharing information between different AI tools so that they're all able to communicate with one another that really levels up the power of Aentic AI because you can bring data in from anywhere and then have this standardized system operate on it.

So then over on the side here we have the history. For example, we have this impact of agentic AI, which is a slideshow that I created last time. So, this is the first example use case is that you're able to actually build presentations. My prompt was generate a PowerPoint and teach me about the impact of Agentic AI and knowledge graphs. And then it went through and created a bunch of refinement questions so that the deep agent could go and create the PowerPoint in a way that's going to best align with the vision that I created. So this is just a good example of document creation where it was able to go through and create really detailed architecture explaining what an AI agent is and how it connects to knowledge graphs. So that's pretty cool. Then I can export the presentation. And again, if you want to see how I did this in more detail, I recommend watching the what is chat LLM video because I go in more depth on the presentation side of things.

I want to get into a few more specific use cases for deep agent here. Now, I just want to quickly give you some rapid fire examples. You can build presentations. We can do research and generate documents. You can have it go through and operate your browser, for example, check the flight prices for you on a trip you want to do and then give you a report by email to show you how the flight prices are changing. We can build in AI workflows where, for example, perhaps it will check every time I upload a YouTube video and then it will automatically generate Pinterest, Instagram, and Twitter posts to promote that YouTube video. It can generate videos for you. You can do data analysis where you're dropping in spreadsheets and having it create reports for you and graphics. And again, we can connect to all these different tools. So, the data sources that we're providing are not just from what we're uploading, but can have this dynamic connection to other services. You can organize your documents in projects like I showed you in the last video. So, there's just so many things you can do with this.

But one thing I wanted to just quickly show you here before we get into the actual examples of today is that you can see that this used up 3,000 credits for a very thorough and detailed presentation. So that's one thing to consider is how much credits are each of these tasks, each of these systems taking up. And what's nice is if we go and take a look at the featured apps here, which there's a whole bunch of different kinds. If you click on an example, it gives you the prompt here and it gives you the estimated credits on what it would take to build that app or build that tool or generate that document. So, if you're not quite sure how much effort a particular task is going to take, you can click on the examples here and you'll get a better idea on how many credits they expect to use up for that particular use case. I'll explain more about the credits at the end.

Okay, so now why don't we build an application because that's one of the key uses of deep agent. And we have two different options for this. The web app here and we also have you can see in the bottom here the AI assisted code editor of the deep agent desktop. Deep agent desktop is more advanced mode. The way I think of it is that the web app is a great place for prototyping, for experimenting, for getting your proof of concept working. But if you want to go deeper into the code and start to manually customize everything yourself, you're going to want to get into the deep agent desktop. And I'll show you more how that works in a little bit.

So building applications in Deep Agent, it's a really great place for beginners because you can just explain what you want your vision to be. And Deep Agent will go through and ask you questions to help narrow in your vision so that it can build something that aligns with what you're actually looking for. It's good to be thorough here. So here I'm going to start and just explain what I want in this app. So I want a rag system for uploading files and I want to build a web app that has this feature. But the cool thing about Deep Agent is I don't just have to build an app that does one thing. I can add a few features here and then give some more context. So I also want to have a backend system so that the data is stored over time. And this is something that's pretty cool too because deep agent lets you actually deploy a backend and a database which means that you can host all of this within Abacus AI and you can build the front end, the back end and the database system. So I also want to include an authentication system so that I can share the app and information with others and verify that they actually have permission to access the documents. And then just considering the example here, Stripe integration for adding paid services in the future. So these are the main features that I thought might be cool to have in a web app. An ability for me to upload files and chat with them, a backend system so that I can save the files I'm uploading, an authentication system so that I can share with other people, but not just make it available to everyone on the internet to control who can see it, and then the option to add paid services in the future.

Okay, so now that we've outlined the features, I've given a couple comments on how I want it to look. Let's click go. So now we can see the deep agent is thinking here. And right off the bat before building the rag web app, it's going to give me a few clarifications. So this is great because in some systems you can go tell an AI to go do something and it'll just go and start doing it and then it'll run into blocks or it won't work in the way that you expected because it made a whole bunch of assumptions that don't actually align with your vision. So it's nice that the deep agent asks clarifying questions before it starts building. And a key aspect of this is it's going to conserve you credits in the long run because it's clarifying on the direction it has to go before it goes and does it. So that way it doesn't start building a bunch of things that you don't need.

Okay. So let's answer the questions. Then I'm going to say let me know if you have any more questions. And what's cool in my opinion is that they're able to actually build the app using the builtin SQL database and host directly on deep agent. So what's going to happen is it's going to give me a subdomain. And what that means is once this is built, I will have a fully functioning application that I can share with people right away because I don't have to deal with deploying the app myself to a hosting provider, setting up a backend, setting up API keys, setting up API routes, setting up authentication systems. The deep agent's going to go through and do all of that for me. Then it wanted me to sign into Google Drive, but I'm just going to tell it that we can skip that connection for now. And here we go. It's actually just right away jumping into starting to build this rag application. Document management and AI chat. So this can take a little while. It depends on the complexity of the app that you're building. So I'm just going to let this run. You can see that as it's going, it's telling me which parts of the app it's beginning to build. So it's already established the authentication system. It set up the file storage. It just set up the LLM. It's cool to see its thinking process as it goes.

Okay, here we go. So now it's beginning to search the internet and you can see its progress as it's going here. It's doing research on what the best way is to set up these different schema, the data style organization for how the LLM is going to query the information, how to set up the history for the document management system. So not only is the agent able to do the code itself, it's also able to go and connect to the internet and run self-defining Google searches based on the task that I gave it. So again, this is where Aentic AI really starts to become a lot more powerful because it's able to go, for example, ground its answer in Google searching to understand the state-of-the-art 2025 best practices, then take that knowledge, build its own context, and then go and code the app directly. And I want to note here too, you can see that there's this main task happening on the bottom here with a bunch of subtasks included. Within the basic tier of Chat LLM and Deep Agent, you only get access to three tasks per month. And if you want to upgrade to get more tasks, you can upgrade to the pro tier where you get an unlimited number of tasks provided you have the credits to execute those tasks.

So once this is finished, it's going to give us an indication on how many credits it took to build this app, which to be honest, it's quite complicated. So I'm expecting it to take up a decent amount of credits, but assuming it works, it's entirely possible that this would have just saved me maybe 200 hours of coding. So I think it's always important to weigh the benefit of using Agentic AI. against the actual time that it would take me to build a system like this. And now it's going through and it's actually writing the architecture documentation on how the app should work. So that's cool that it's actually building a document to explain how the application is working. And this is where again it's nice to have all of these features in one place because for example I could then take this document and create a presentation for marketing materials to go help promote the app that I just built here. So by having all of this happening in one place within this web app, you get a lot of power to have the agentic AI do various elements, very multimodal creation. And again, a major benefit of this is it's all happening autonomously. So I can set this up, I can have it run, and I could go make lunch or go for a walk or start working on something else and the agent's going to continue to operate in the background building the app for me.

If we click up on the top here, we can see the files and we can see that it has the rag app here. So I can just download the app which is great because that means that I can take this and start working on it within the desktop application for example and just continue where I left off or start using VS Code. You can just continue building on the application once it's been made in this app here. It's already coded all the different components of the application. Now it's just going through and setting up the connection between everything. And all this has taken about 5 minutes so far. So that was just the first part of the major task where it went through and figured out how to set up all the database systems and it created a 65page document on the optimal design for the system. Even if I have to go in and start making some tweaks and iterating on this, not only will I have the basic app built, but I'll also have a ton of research that I can use to enhance the application myself afterwards with very targeted enhancements. So this is the output that it just gave me here. all the different documents here, how it should build everything, how the different databases are going to connect, and then also creating a checklist for itself on how it should go through and then deploy and build this. So, this is great because if something happens and it crashes, for example, I'm able to just take this document and I've already done the research for setting it up and I can just continue where I left off. If I take this to the desktop app, which I'll show you in a few minutes, I'll have additional context that I can go in and make the targeted fixes or enhancements that I'm imagining here with a little bit more of an advanced mode.

Cool. So, it just finished its research. It created the comprehensive engineering guide for document processing. It's actually updating the application itself. You can see the files specifically that it's updating here. Oh, and it looks like we now have a deploy button that has appeared and it's going through and you can see it's actually building all of the different files that it just planned out ahead of time. So, this is where Agentic AI can work to create a plan and then execute on that plan all on its own just based on that initial prompt. So, it's been going through and autonomously establishing what it needs to do in order to actually build this application and now it's going through and actually building the application. A reminder to please like, hype, and subscribe if you're finding this video helpful. Your support is very much appreciated. Also, if you want to check out Deep Agent, I do have a link in the description below. Now, let's keep building.

You can see it's actually running the application in its own little virtual environment so that I can go through and start testing the application itself. So, once the app's created, I'll be able to preview here and then I'll be able to go through and start viewing the logs for debugging. I can deploy the app to my own subdomain. I'll show you all this in a minute once it's done coding, but so far I'm quite impressed with the level of detail that it's gone through in this process. I've used a bunch of different agentic AI systems and rarely would it ever use more than one or two Google searches for example. But you can see here it did like 15 searches for each component of the application trying to understand how it can build the best app possible. I'm excited to see what it looks like. So as it's going through and doing this, it keeps running different elements and installing components into the application. You can always click and see the command output. Okay, here we go. Looks like it's actually starting up. Okay, here we go. Here's the preview. So, that took about half an hour exactly to go from start to finish with building this. And honestly, given the amount of files I created, I'm very impressed. So, let's just take a quick look here. You can see it says called rag app, which is uploading, organizing, and chatting with documents using cutting edge AI. Turn static files into dynamic knowledge with intelligent conversations. So, my thought is I could take something like my Obsidian Vault and connect to this rag service and now I have the ability to connect with my note-taking system directly. That's pretty powerful. It's even got the ability to hover. You can see the UI looks pretty nice. It'd be nice if there's a dark mode, which maybe I'll take a look at adding in a moment here, but you can upload multiple formats, AI chat, smart collections, so you can share with people. Honestly, this looks pretty good.

But let's take a quick look at what Deep Agent just gave me. We can use the deploy button to make it publicly accessible. So right now, this is just a preview that's running within Abacus AI right now. So if I click this deploy button here or here, it will actually deploy the website so that anyone can access it and I can share with people and start using it right away. It gives me a bunch of steps that I need to do in order to enable the signin with Google. So for example, if I click sign in right here, you can see it has continue with Google or continue with email and password. Oh, interesting. It's already pulling up my iCloud passwords, too. Cool. and it gave me a test admin account. So, why don't we try that? And just before we see if it works, you can see here this used up 3,330 credits. That's a significant amount, but I want you to take a second to understand what we actually just built here. So, there's a full retrieval augmented generation system, which means I'm able to upload any type of file and then chat with an AI. All the files I upload get stored in a new database that operates with its own backend. So I can query the data that I have stored and it persists across sessions. So I can refresh the app, I can close it and it should still work. There's a full authentication system in here so that for example I could give authorization to only people who provide their email address to sign in to get access to it. And we can start to potentially have different collections of information and data so that people can join a team for example and start chatting with the documents. And then I also built in a Stripe integration so I could add a paid service. So, for example, maybe you can only upload a certain number of documents before you have to start to pay. And all of that is integrated in here with what Deep Agent was able to build in half an hour. That's pretty mind-blowing. I mean, of course, we have to see if this actually works here. So, I'm excited to test this out. But I've used a ton of different agentic AI systems. I've used Claude Code. I've used Chat GBT. I've used Client. I've used RU. I've used Gemini CLI. And this is the most comprehensive singular creation process that I've seen with an Aentic AI coding system. The fact that I was able to build this entire app all in one go, that's mind-blowing, but let's see if it actually works.

So, this is the homepage. Looks like it's got a full profile. Some of the features aren't activated yet, like the settings for example. Okay, so let's create a new collection. Let's call this Wonderloot Tutorials. Cool. Okay. So, I can set this up as a private collection so only I can see it or I can share this with other people. Let's click private. Cool. So, you can see here there's a new collection on the side here. I just clicked on it. Let's upload a document. Oh. All right. So, here we got an error. So, I would say so far the UI looks great. I was able to create that new tutorial. I could upload documents. I can search my documents here. And I can save chats and have chats. And obviously, there's going to be a few different bugs here because I just built it and I'm just testing it right now. But so far, I'm able to have the full authorization and sign-in system, which I've built apps, and I actually just set this up on an app I'm working on now. And that took me maybe two weeks to get working properly with my application with Cloud Code and Google AI Studio. And already we have the authorization system working here. That's pretty cool. Obviously, it depends on your stack. And my stack was a bit different than the Nex.js that I'm working on here. But all right, I click on this, and I get an error.

So now let's see if I copy this. So here we go. I just gave it some feedback and I said I got this error when uploading the file. Can you please debug? So just before I deploy the app, I want to see if I can fix that error just so that we actually can test the flow from start to finish. But this is definitely a core component of building with Aentic AI. I just said, I got this error when clicking upload file. Can you please debug? And it went through and it said, yep, here we go. This is the issue. I had a bit of a UI and now it's going through and it's debugging and making the fix directly. One thing I want to point out is that these things aren't going to work perfectly from the beginning. There's going to be some issue that happens at some point and the key here is to know how to work with those issues and fix it. So, for example, I just asked it to debug and I gave the error code that was explaining what was going on. The deep agent went through, found the problem, implemented the fix, and now it's running it and then testing it to make sure that it's working. So, it's honestly really nice that it's running the testing at the same time. I've had to do this manually myself with other AI tools. I can't really overstate how helpful it is to have the full test process happening autonomously. Like, this is saving me so much time already.

Okay, so that took about 2 minutes or so to debug. And let's see, I'm still signed in. Let's see if it works. There, the problem was fixed. And one of the most helpful elements that I find personally of using Agentic AI for coding is that not only does it go and make the fix for me, but it's specifically explaining what the problem was and how it was fixed. This is a great way for you to not just build apps with AI, but learn how to build apps properly where you can start to learn how to code, learn how the pieces fit together, and you can really start to get better and better as you experiment in these types of systems to make better apps. Let's see. And that debug, for example, used up only 240 credits versus the 3,000 credits it took to build the initial app. So, as part of the pro tier, you get 25,000 credits. I would have used up roughly 15% of my credits so far to build this app.

Let's see if it works. I'm just going to upload the PDF that it created for me. Okay. And it looks like we've got another little bug here. If I click anywhere in this bubble, it's doing the same thing of just upload file. So, here's another bug we need to fix. Oh, interesting. So I asked it to debug and right now it just created a virtual version of itself and it's actually going and it's autonomously actioning the computer for me. So this is actually a great demonstration of what I was talking about with Agentic AI where it's able to go through and actually operate this virtual computer for me. So you can see this is a virtual workspace. It just went to connect to my local computer and it's operating Google Chrome right now for me. There we go. So, by operating that virtual computer, it was able to go and identify the issue. And now it's going through and fixing it. It's kind of surreal watching it go through and click in the virtual computer and actually operate the app itself. Oh, there we go. See, it does work. It was able to change now to web URL, which I wasn't able to do before. And it's even able to go into the developer console itself to figure out what's going wrong when things go wrong. So, that's pretty cool because again, this is all happening fully autonomously. So, I can just set it to go debug. So, I don't have to keep going and testing it myself. it will go and autonomously test for me. Perfect.

Okay, it just restarted. Now, let's let's give this a go here. I'm going to upload the document again. I'm going to try using a web URL. Bringing in one of my newsletters. Process it. Perfect. It just uploaded. Okay. And we can see that used up a little bit more credits because it was a bit more of a complicated fix. So, you have to be careful here how you wanted to debug. For example, I could ask it to tell me how to test things to see what's working and what's not working rather than have it go and do all the screen recording, which took a lot more credits. I hope you get the idea here. So, I just asked a question there on how do I actually start a chat. It had to rebuild the app once it gave me the answer to the question. So, that could be a bit inconvenient where I want to be able to ask questions about the code itself without having to rebuild every single time. And this is where the desktop app comes in because the desktop app lets you actually chat with the code itself rather than just running the agent on everything. So I'm just going to quickly show you how we can get this to deployment and then I'll pull up the desktop application to show you how we can continue debugging in a bit more of an efficient manner.

So once I click the deploy button, we have a few different options here. We can have an abacus.ai domain. We can input a custom domain and we can use a custom subdomain. Now, that's a little bit more complex and I don't want to get into that right now. I do have a video on how I publish my Obsidian Notes website for free using a digital garden. And in that video, I walk through how to set up a custom domain for your own website. So, I recommend checking that out. But for now, let's call it chat with Docs. Cool. There we go. It says it just deployed. Let's click done. So, let's go to the website. Now, anyone would be able to go to chat with docs.appacusai.app. We can sign in here and let's see. Perfect. So, the app is actually running and this is where it's been deployed. There's a full back end operating and honestly that's pretty great. That's impressive. Obviously, there's some debugging here where I need to get the chatbot working where I'm actually able to have a conversation rather than just using the Deep Agent web app.

Let's take a look at the Deep Agent desktop app because it might be able to be a little more efficient with how we debug some of these targeted fixes. The Deep Agent desktop app gives you access to AI coding in the editor itself. So, I'm going to click sign in. Dark mode, of course. We have two options here. We have code mode and chat mode. And it sounds to me like the chat mode is basically what we just did. And we can always toggle between code mode and chat mode up here. And you can see it's actually a version of VS Code because it's pulled in all of my VS Code extensions here. So you'd be able to download your files like we had that zip folder before or you can connect to GitHub and clone your repository. And we can go and start making all the changes that we would want to inside of the desktop app. But that's going to take me a little more work to get set up. I just want to show you that this is possible here. So you can start to go in and you can chat with the agent on the side. So, for example, if I want to debug the app, I can clone the repository here and I can use different models outside of just the deep agent here to make those changes. And I mentioned that I'm not able to change the code here because it's in read mode. But there is actually an edit button right here. So, I can click edit. And by doing this, I'm able to now go in and find specific files. So, I could open this up as a separate window. And I have the preview application running right here. And I could make changes and see if it fixes everything here to have that chat operating. And this would continuously update and maintain that deployment at chat with docs.abacusai.app. But back to the desktop now. I can go like this. And for example, we have the rag app here. So I can just click download the zip file. I can go over to my abacus app here. And I'd be able to open a folder here. Go to my downloads. And here you can see that I have the full application here. All the files, everything was just imported directly into my application here. and I can have a conversation with my documents here with my code and see if I can go through and debug and make sure that everything is working. So, this operates the same as a normal coding editor. So, I don't I don't need to spend too much time here. I just wanted to show you that you're able to change the model here, whereas in the web app, you're not able to do that. It's going to just use whichever model it thinks is best, which is probably the deep agent model. Obviously, again, I've got a few steps to debug, but you can see it just took a couple minutes for me to debug that one feature. So, I'm going to go through and debug that and show you how it works in a moment. But, I hope that gives you a good overview on how this all works, how you're able to go and have that conversation with your app here and then iterate and improve on it over time to continue building whatever it is you want and have that full deployment capability.

Okay, so I took some time to debug. I added a few more features and I thought I would now show you how the app looks. You can see here I've added in the dark mode. So, I can now toggle back and forth between dark and light. I cleaned up the interface a little bit, added a search bar. I introduced a chat history section and also how it works section. So, if you're interested in learning more how this app works and what the full architecture is, you can just go to this URL and sign in. There's a lot of information on how chat bots work and how you can kind of understand the different elements that went into building this, including the front end, the O layer, the rag engine, the API layer, database, and file storage. So, it's pretty cool that all of this could be built just with one application. And why don't we see how it goes? Let's upload a new document. This is one on the Fman technique that I made a video about. And I want to see if I'm able to now chat with the notes that I made for it. Cool. There we go. So, it uploaded no problem and embedded it. It processed it and now we can have a conversation. There we go. The AI chatbot works. So, I can ask questions about it and use the technology of Rag. So again, if you're interested in learning more about all these different technologies, I created docs using Deep Agent that explains how this all works together and you can explore this if you're interested. But overall, I think this is pretty cool that you're able to build and deploy the app, both the front end and the back end, all using Deep Agent. It's also cool that you get this subdomain for free as part of your subscription. Normally, if you wanted to host your app, depending on the provider, it could cost you anywhere from $3 to $15 a month. So, as long as you're paying the monthly subscription for Abacus AI, you get access to Deep Agent and the hosting of your apps. So, this just lets you share with other people so they can go test it out. And altogether, I would say this probably took me about maybe 10 more attempts of debugging with targeted fixes. So, that used up probably around another 2,000 to 3,000 credits. Altogether, this took up about 6,000 to 7,000 credits, which means it used up roughly 1/3 of my monthly subscription. And honestly, that's not bad given how many features I was able to build into this application. Like there's a lot that you can do with this and honestly I'm pretty impressed.

So now that you have an idea on how all this works, let's take a quick look at the pricing. So the key here is that it costs only $10 a month to give you access to Deep Agent Chat LLM and Deep Agent Desktop. But with that, you only get three basic tasks of Deep Agent. Maybe that's good enough for you. Maybe you're working on presentations where you just want to create a couple presentations a month and $10 will get you quite far with Deep Agent. But if you're going to be doing a lot of coding, probably are going to have more than those three tasks. You're going to want to upgrade to the pro tier for an additional $10. It also gives you access to 25,000 credits instead of 20,000 credits. And it gives you access to that more powerful version of Deep Agent that I showed you today that was able to autonomously do all of this. You're going to need to have some experimentation here to see if you can get it all working to make sure that it suits your particular use case. But if you're a professional and you're using this for coding, I could see needing the pro tier in order to access the full features and build everything that you're looking for. But if you're using this for more basic tasks, you're probably fine with the $10 a month. I think being able to operate in this web app is a great place to experiment and get the MVP going. I hope that this video shows you how powerful using Agentic AI like Deep Agent can be for document creation, app building, and doing everything all in one place. If you found this video helpful, please like and subscribe as your support really goes a long way. If you want to support me further, please consider joining my membership or watching some of my other videos like the AI learning playlist where I talk about a whole suite of different AI tools that I use in my day-to-day. Thanks again for watching and I will see you in the next video.