How to Build your Own Agent Operating System!

Key Takeaways

Today, I want to guide you through exactly how to build your own agent operating system based on what I've built out. Now, just to inspire you and show you some examples of what we've created, we've got like an agent orchestration section here where we can have teams of agents working together. We've got a agent group chat where we can see the agents working together. We can orchestrate them that way. We can also, for example, have a full pipeline where they take the ideas from the group chat, drop them in the pipeline, then we can ask it to just build it, and we can see everything that we've previously created, as you can see right here, which is pretty wild. Then, so. Then, we have all of our CLIs and agents plugged in so we can use them separately, see what they build, preview stuff that they've created, etc., which is pretty cool as well. And, you know, for example, if something new comes out, like for example, GLM 5.2 drops, that's kind of like close to frontier level, and you can use it on the coding plan with your AI agents. You can see everything that we've created over here, and it's quite easy to just build stuff and then preview it later inside the workspace, right? You can also see like some other cool stuff that we've built here. Um and the quality is actually amazing, but also it's quite easy to just build this stuff out and have it inside one place. And also, the great thing about this is you don't need to have like separate tabs or separate terminal windows. You don't need to worry about coming back to your setup or, you know, let's say for example, you shut down your computer and then you want to come back to it later. Well, if you got everything inside one agent operating system, that's going to save you a lot of time. We also have a memory system over here. Again, just to show you and inspire you in terms of what you can build here. So, this is a memory galaxy. We can see everything we've built. We've got Kanban boards with teams of agents working together. Notebook LM plugged in so we can see all the stuff that is built over here as well. We even have a local system here where we can speak to our AI agent. This is locally set up with GPT RSS. And then also, we can see everything inside the workspace here, too. Then, we've got Fusion, which is like a powerful way to run all your It's kind of like a powerful way to get higher levels of intelligence that just dropped, and we just plugged it inside here, as well. So, like everything we could ever imagine is built inside this system. I'll even show you for example another example here. All right, so this is a video agent and we created this with a combination of all our workflows and then just built in a video agent where we can type in a a prompt, one single prompt. This is the prompt over here. It researches it, writes the script, edits it all together, adds a B-roll, puts in an AI avatar as you can see and it looks absolutely amazing. All right, it's all put together and and properly synced together. So, this is an example of what you can build and how powerful it is. Now, let's talk about how to actually build it and where I started and how I got this all working together. So, you know, you can see examples already of what we've built and how it works together. And, you know, if you're tired of like having 14 AI tabs and no idea which one to choose every session, re-explain your business from scratch, your subscriptions in loads of different places you can't even find, losing files, losing contacts, starting from zero pretty much every single morning. Well, that's basically the problem that I was having and that's why we rebuilt this, right? And so, we built one screen that runs everything on one shared brain. And now, after this I open one dashboard, every agent is there, they all read my Obsidian vault, I can hand them all one sentence, if it comes with a finished build ready to go and we can have agents working in parallel with the orchestration systems I built was ready and you can see examples of like what people are building inside the Air traffic warden community using my exact same setup. But basically, you've seen how this works, now let's talk about how this works and just promise yourself one thing right now. Like, you finish this and you install one piece of it today because when you do, when you commit to building, you're going to see how powerful this is and and what you can actually do with it and [clears throat] anyone can do this, right? I'm not a coder, I'm not technical, but I could do this as well. So, how does it work and how should you think about this? So, you want to think about your agent operating system like this. You ask something, you have a shared brain and then you can plug this out to Claude or G Lam or a finished build, right? And so, it's six layers stacked on top of each other, and you don't need to have like G Lam and all these different CLIs. You can just have one single CLI plugged in, and then use that, you know, or you could have one agent plugged in like Hermes, whatever you prefer. And so, the way to think about this is like six different layers, and the way to get started with this, this is how I do it, is I I'll go into for example, Claude here, Claude desktop, and I can start building out the agent OS, right? So, I just describe what I want, and you can just start with one workflow. You describe what you want, and you say, "Hey, build out a mission control for this. Start doing this." And that's how you get started. And then it just stacks on top of each other until you got a full agent operating system. So, you get started, and you just say, "Create a mission control. Here's my idea. I want to start building with this." Now, I improve this every single day. I love working on it. It's one of the things that I spend a lot of time on. Um and also, it's super useful. So, that's the first place to start. Now, you want to think of this in terms of, you know, when you're prompting Claude build this for you, you want to think of this as like you want one screen for everything. So, a local dashboard that holds every agent, every tool, every output, and you stop hunting through 14 different tabs. You just have everything living in one place that you actually control. And so, it all lives in one place. That's the first layer of this. How do you do that? You just go to Claude and you say, "Hey, I want a mission control. Here's what I want you to build. Create the dashboard for it." And you just set You set it up, and you test it. From there, you also want to have a brain plugged in. So, a memory that never forgets you. Now, for me personally, I have Obsidian wired into this vault. Let me show you. So, if we go inside the memory here, you can see an example of Obsidian. If we click on any of these memories, we can actually view the recent memories here, what we've built, etc. And if we click on something like this, for example, we can scroll down, and we see all the notes for that particular memory. So, Obsidian is a free app that you can download, run it locally, plug it into Claude, and all your agents that live inside the agent operating system, and that's step number two. So, step number one, you tell Claude exactly what mission control you want to build. That's the actionable bit. Step number two, you're going to plug in a brain. Step number three, you start stacking in more models side by side. So, for me personally, because I'm always testing things as I have to, and staying up-to-date with everything, I've got Claude, I've got GLM, I've got Kimi, I've got Hermes, each plugged into the same brain. And so, for example, what that means is that I can orchestrate them, and I can use cheaper models for the more time-consuming stuff, and I can use more frontier models like Claude for the plan. And if I have a good plan in place, it doesn't really matter which model I use, because they can implement them very well all together. And then, you also want to have a workspace, where every build is saved and previewable. So, if we go into, for example, with GLM here, and we see what we've built. Here's an example. We can open that up from our workspace, and we can start playing with this now. All right. And so, you can see how we can start playing this, because it's all saved inside a workspace. So, we can come back to it. We can see everything that we've built, and that just stops you losing everything, and also it plugs into your Obsidian system. So, if you want to like improve it with a different CLI, you can later. And so, everything is saved in one place. Perfect. That's a workspace. And then, also always on. So, I have this running locally for two reasons. Number one, I can just come back to it whenever I want, and number two, it runs 24/7 without me. And so, it's working in two ways. Number one, it's always building stuff without me, and number two, it's ready to go whenever I need it, which means like, I don't have to go over to Claude, and then open up a new chat, and click over here, and then think, "Okay, where am I? What am I doing? What am I working on?" No, no, no, no. Everything is ready to go right here. If it's like, "Do you know what? I want to ask my team of agents what they think about fusion and the new AI that just came out?" I can ask it right here. If I'm like, "You know what? I need to create a new video today." Okay, let's preview it over here. Or it's like, "Okay, I want to see stuff I've previously created with my video agent system." Okay, great. It's over here. I can come back to it later. You know? And so, you have everything ready to go whenever you need it. And then, the other thing here is the loop. So, this gets smarter every single day. And I'll show you an example. So, like every single day, I'll have new ideas or I'll be able to build on top of stuff in terms of what I've created. So, if we have a look, for example, here this morning, this is just this morning. Been working on setting up local models inside this, setting up Hermes with local models, setting up the Hermes agent loop section, which is a new uh loop section that we can build anything with, and also setting up free code codes to make it even better. Now, that's only because I've already got the foundations in place and I've got ideas of how to make the whole system improve. And so, every day, I'm excited to work and build on this because it just gets better the more I use it. And the more I build with the memory as well, the more the rat memory becomes useful as well. So, just to recap, how do you get started? You go into Claude, you tell it exactly what you want to build. What are the most useful features? The workspace, the memory, different agents plugged in, and the agent orchestration. And also, just having everything ready to go 24/7. Now, why would you use this over the old way? Let me talk you through that. So, the old way is that you might waste like 2 hours a day because you got like 14 different AI tabs open, none of them talking to each other. You have to re-paste your business context in. You're using different subscriptions separately that you can't even find. You have one model, one job, and you just have to wait for it to finish before the next model runs. Your output's scattered across different tabs, half of them lost. You close the tab, the AI forgets you completely, and the result is generic answers, lost work, and starting from zero daily. Everyone's experienced that. Everyone that I know has experienced that. The new way is a maybe in 20 minutes a day you got one deck, so one mission control, every agent on one screen. They read your vault, your Obsidian vault, so they know your business before you even type. The free agents and the local agents can run the grind, and you only use frontier models where it actually counts. And then you can have multiple agents working in parallel whilst you do have other things. So, for example, if you go inside the system here, we can have multiple agents working in parallel inside the group chat. We can also have them working on the pipeline. We can also have them working inside paper flow. So, we can have a AI agents working in parallel together whenever we need them, which is absolutely amazing. And then the cool thing about this is every build is saved and previewable in one workspace. You close the laptop, the brain remembers everything, and the result is answers grounded in your work, nothing lost, and it all compounds, and it all improves every single day. And you might say, "Well, this is just for tech for coders or technical people. I'm not technical." Well, if you can paste one command and click open on a folder, you can run this. If you can go into coding and just say, "Hey, here's my idea." you can build this. And if you want the whole system or you want to see other people building this, not just me, here's an example. So, you can see all these people like Amanda, she posted about the agent operating system that she's built. Ken built an amazing operating system. Marion built one as well. Like they're all using the systems that I've built inside the AI Profit Room, and they just implement it themselves. And so, like if they can do it and if I can do it, anyone can do it. Now, how do we actually build this step by step? So, here's the real order that I built mine. Each stage and each step is one layer of the framework. You can stop after any step and still have something useful, but each stage makes the next one more powerful. So, step number one is the dashboard. So, we started with one screen, simple local dashboard that every agent and tool plugs into. You can just ask Claude to build that if you want to. You don't have to hump through tabs anymore. You can just open one window, and the whole operation is in front of you. Every agent down the left, every output in the middle. So, that's step number one. Now, step number two is you wire in your memory. So, this is the piece almost nobody does, but it's a whole game, which is you want to point every agent at one Obsidian vault. Your notes, your clients, your goals, your voice. And now, when I ask any agent anything, it can understand me real fast, and also it logs all my information inside the memory every time I use my agents. And so, it's just building and improving all the time, which is amazing itself. Now, you might also might think, "Well, I already have a tool. Why do I need this?" The thing is, the Deck runs all of them, including any tools you already have. So, it doesn't replace your tools, it orchestrates them. And it gives them one shared memory, so they finally know about you. Additionally, all your different workflows that you work on day-to-day, we build inside there. So, for example, SEO, we've got a section for that. Video, we've got a section for that. Even a music agent, and I actually regularly use this, as you can see. Uh a game studio. We can generate images. We can generate podcasts and voice notes and videos and everyone. We've got a Kanban board, where our agents work together. We've got the memory set up. We have local models inside the local section, and we can see everything that we've built with our local models, too. So, this is just a powerful way to see everything inside one mission control. And then you have the crew, which is you can plug in every model. So, I started with Claude, and then I added GLM, Kimi, Hermes, and local models as well. And each one sits on the same brain. So, you pick the right worker for the job, like a real team. The screen show below is one agent's panel, and you can just chat with it, build with it, preview what it does. And you don't need to use all of these, you can just pick one if you want. But you could have multiple agents later if you want to, as well. And then you got the workspace, so you never lose a build. Like, I don't know about you, but how many times have you faced a situation where you basically build something, and then you can't find it later. That is the situation I've seen over and over again, and that's exactly why we built this. Also, it's always on, so you can make it run without you. Now, I set the whole deck to boot a login and it's pinned, so I can just come back to it from anywhere. You could also connect this to your phone with Tailscale. I wouldn't recommend that because it has access to your computer. I don't think that's secure, but you could if you want to. You might also say, "Okay, what if it breaks?" I I've broken it many times, uh to be honest with you, but you can always get Claude to fix it. It's pretty simple. Usually Claude saves checkpoints as well. Every fix that I make is put inside the AI Profit Boardroom, plus the community uh actually direct message me if anything is a problem and then I fix it for them as well. Now, the loop as well, this is the last layer. So, every chat also logs in to the vault. So, tomorrow's agents know everything today's did. And that's the difference between a tool and operating system because a tool is the same on day 90. The deck is smarter because it's learning and it's proven, and I'm building new stuff into this all the time. And it can work for anything like SEO, video, apps and games, client work. Like, I'll give you an example. So, we wanted to start going from like idea to generation really quickly. So, we created the ideas list here. And so, we can plug in an idea and then it just builds it, we approve it, and it actually gets built, as you can see inside the previous section here. Pretty amazing. Then we got the research and the automation in the background. Again, some people think, "Well, I need to be a developer to build an Agent OS." You just need to talk to Claude. You don't need to be a developer. You can just talk to Claude and tell it what you want. Some people say, "Well, I'll wait until someone makes this super simple and easy." I don't think it's coming at this point, but my Well, my whole setup is inside the AI Profit Boardroom. So, if you want my setup, it's inside there. If you want to build it yourself, you can. If you're waiting for someone else to build it, I don't think that's coming. You know, otherwise it would have been built by now. It is just such an obvious idea. I don't know why people have not publicly put it out there with something good and reliable. But also, it takes a lot of time to build. That's the other thing. You know, it can take hours to make one single improvement. Like, for example, the the local system that we set up earlier today, that was just like 4 hours of building and just having Claude constantly working. So, it's something you have to be aware of. It's like it can be quite time-consuming if you're building it all yourself from scratch. Enough people say, "Well, AI tools are fine as they are." But honestly, if they're not plugged into a memory system, if you don't have your workflow saved, if you don't save everything that you create, well, you're not getting the full power of AI and you're missing out on 99% of the stuff that you can actually do with this. So, that's basically it. That's how to do it. We've actually got a full 30-day roadmap as you can see inside the AI Profit Bot room. And what you've learned today is you learned how to stop juggling loads of tabs and terminals. You learned how to stop re-explaining yourself to your agents. You learned how to use free models and local models to build into this stuff. You learned how to stop waiting and just have agents working in parallel with the orchestration agent systems that I showed you. You learned how to stop losing work by having every build saved and previewable on one workspace. And then you learned how to have a positive feedback loop so that it just improves every time, right? Now, if you want my exact deck and you want to set up with me as well. So, any improvements that I get, people DM me all the time and I just implement it for them same day. I release new daily updates inside our community as well. And we also have a full system that you can just install from a quick zip file. So, if you want the agent operating system, you can get that inside the AI Profit Bot room. It turns Claude, um Kimi, Hermes, whatever you want, into one deck with shared memory on one screen. You can control via agents understand you. They remember everything and every new AI tool that ships makes the whole deck more powerful automatically. So, you get the full agent operating system zip file in there, coaching calls, the 30-day roadmap. I'll show you an example what this looks like. So, if you go inside the classroom here, you can see the full agent operating system build like so. You can get the video tutorial. You can see it's just updated today. You can get the zip file as well. You can ask questions inside the community and I get back to you ASAP as well. So, you can see here that I've answered all these questions. And then also inside the calendar, you can jump on weekly coaching calls, get help and support real time, share your screen, etc. Inside the map, you can meet people in your local area who are building out with AI agents like you. And this is all inside the AI Profit Forum. Link in the comments and description or go to the AI Profit Forum.com. Thanks for watching.