I Built My Own AI Operating System With Claude

So today we're going to run through some of the latest AI questions that we've had first of all. And if you're watching, you have any questions, feel free to ask. I am happy to help you. And we're just going to get straight into this. So these are questions from the AI Profit Boardroom community. So question by Issyaki, which is VPS for AgentOS. Why do the pros avoid cloud hosting? So Issyaki was saying, you know, read the community Q&As daily. Notice the trend. You often mention you prefer running AgentOS locally. Instead of VPS. What's the reason for this? Is it strictly for data privacy? Is it too slow? So let's get straight into this. So question number one, personally, I prefer to use the local hosting instead of cloud hosting for an AgentOS, simply because I think like it's really good for security, right? Like if you look at most of the problems that happened with OpenClaw when it was first released, it all came down to having VPSs, right? Like there's people actively searching for VPS and people actively like looking for access to VPS. And so I don't think it's the best way to run an AgentOS. And also bear in mind, like you don't need a powerful setup. So like even if for example, you just have a normal laptop or even a Chromebook or something like that, you can use AgentOS locally because it's just like a web app, right? So when you're building out with this, it's super lightweight. It's just like a website. And so you don't need a powerful setup to run it locally at all. Then second question is, many of us are running like a micro business. At this stage, enterprise level data isn't the bottleneck. Agent volume is fairly low. So it seems like a practical solution. So if you run an AgentOS on VPS and then someone gets access, it doesn't matter whether you're an enterprise or a small business or not. The biggest issue is like someone could actually access your local computer, right? If you basically accidentally give access to AgentOS. So that's not good for anyone, no matter what sort of business you run. And that's why I always recommend like just running an AgentOS locally instead of hosting it inside the cloud. Having said that, some people to run a VPS, but I just don't see any benefit. You definitely don't want to give anyone access to that. You definitely don't want to give anyone access to your local setup, which is where your AgentOS is running. So for me personally, I just, yeah, I'm always going to run it locally instead of on a... So that's basically it for Issyak's question set. Let's have a look at what we've got here. Venn says, I want to create some workflow for interior designers. What sort of workflows you want to build? If you can be a bit more specific, I can probably help you out and give you some good tips on that. Mahamud says, welcome. Hello. Good to see you. Flux says, which AR harness do you use? Good questions. All right. We'll get straight into that in a sec. So for me personally, my two favorites in terms of Agent Harness are Hermes and Claude. I also really liked DeepSeek Harness. I think anyone watching this, the best thing you can do is just focus on one at a time. So if you're not sure which harness to use, just focus on it. If you wanted to get my best recommendation, it would be Claude Code. Everyone has their own preference. So for example, you know, if you like open source, then you probably use Hermes because it's much more flexible as a harness. If you like something that is less buggy and more likely to work every single time you use it, then I would go with something like Claude Code instead. For me, for example, I don't particularly like Codex as Agent Harness. I just find it's like a little bit too distracting. And also with DeepSeek Harness, like when I use it, I just, I feel like the UI is just too basic for me. Like it's not fun to use DeepSeek Harness. As useful as it is, from a UI perspective, it's not that easy to use. Whereas like Claude Code, my brain is fully trained. That's why I prefer to use something like Claude Code instead, which is the best way to start design workflows. So for me personally, there's two things that I like. Number one is building my own design workflow. So for the image generator that I built in the Agent OS, we can easily in the custom workflows I have here. And it creates beautiful images. It's really easy to build. So that's option one. It's like basically option one is build your own workflow. UI, look exactly how you want it, which is what we did with the Agent OS and the image generation section here. Option number two is that you can just use something like ChatGPT image generator. You know, ImageS 2.0 inside ChatGPT is really nice. And if you need to build a custom workflow like the Agent OS, you can just get the API or ChatGPT image 2.0 and then build a beautiful tool like you can see inside the Agent OS. So Maturza was asking, you know, where is the Obsidian setup? So for me personally, my favorite Obsidian setup is with OMI, Obsidian and the Agent OS. And I'll link to that below. This is using a system I call the Infinite Context Engine, which is basically where you can get number one, Hermes Agent or any other agent that you prefer to use to update your Obsidian memory for you. So for example, every time we use the Agent OS, agents update the memory system for us. Then OMI actually records my screen and takes notes on me and what I do all day and, you know, what I'm working on so that basically the notes are automatically updated and I don't have to do anything. And also it has context for me. And then that's logged into Obsidian, which runs locally. It's a free open source app. And this works as a second brain that we can plug into all our agents. So I can plug it into Claude, into Hermes, into Antigravity, whatever I prefer. And they have instant context from day one. So if we open up Obsidian here, this is an example of the knowledge base that we have. So I would recommend starting with this training right here, the Infinite Context Engine Guide. Plus it has a video tutorial on it as well. You don't have to use OMI as well. So just know that you don't have to use OMI if you don't like it or if you don't want to give access to your screen to an AI agent, that's totally fine. You can just, for example, say to Hermes or Claude or whichever agent you use, every time I use you, please update Obsidian directly for me. Well, that's basically the system. But yeah, Obsidian is my favorite setup. Super simple, super easy to use. Literally so easy to build because your agents already have memories that they can plug into the system. It works really nicely. If you have any other questions, feel free to ask Matazza. Thanks very much for joining. Let's see what we got here. So Alexander, wow, this is pretty amazing. Pretty amazing. Congrats, Alexander, on the win. Awesome to see the progress you've made. I think that's pretty much it for all the questions that we had today. Let's see what we got on here. All right, two good questions here. So number one, do you use Obsidian or have an AI brain setup? So I actually think that, I mean, Obsidian is a brain, right? It is a second brain. So that's the way that, how do I use it? So basically, the first thing that I do is inside our organic operating system, I have a system, which you can see right here, where we basically have a memory knowledge graph. And so all of my memories, everything about me is automatically noted by my AI agents, plugged into Obsidian. And then Obsidian is used to train my other AI agents and all the contacts and all the projects I'm working on. So it's kind of like a positive feedback loop where when I work with an AI agent, it notes its memories into Obsidian. And then when I use other AI agents, they can pull that context from Obsidian and use that to give me more personalized, better outputs. But the other great thing about that is when I get those personalized, better outputs, anything that I do from then on onwards and logged into Obsidian. So Obsidian feeds into the agents, the agents feed into Obsidian, and that makes both of these tools more powerful. So for example, every time I do something useful with Hermes, that will be logged into our Obsidian database. So you can see, for example, here we've got a new memory, which is Hermes Astros, the custom workflow we've got, which looks at our competitors and gives me new ideas for creating content. And so that's all logged into our Obsidian memory vault so that we can pull from that for future workflows. And I could take that information and go to Claw and say, hey, create a video about this or create a blog post about this, et cetera. The next question we got was, what do you prefer, Hermes or OpenClaw? So for me personally, I don't use desktop Hermes or desktop OpenClaw at all. What I prefer to do is have both of them built into an agentic operating system like this, where I can build custom workflows. For example, if you look at Hermes, and that's kind of what Hermes Desktop does well. But then we also have, for example, Hermes Bots, Hermes Apollo, Hermes Oracle, Hermes Muse. These are all beautiful custom workflows that I've built specifically for myself, where I can take the full power of Hermes, but I can design and mold it exactly how I want it. Whereas, for example, if I open up Hermes Desktop, it doesn't look as useful or as personalized as this. So that's the biggest difference between all of this, is that you don't want to use Hermes Desktop or OpenClaw Desktop straight out of the box. The problem with those is they're not customized and personalized to you. Whereas, for example, if you have an agent OS, like you can see right here, and we give it to members inside the AI Profit Boardroom, it looks more personalized. It feels more personalized. It's more unique to us, right? And that's way more powerful for me personally. Another question was, did you build the web app? So yes, the whole of the agentic operating system I built from scratch, it literally just started off as like a basic Claude and Hermes as a chat. And then the more that I used Claude Code Desktop to build it, the more I used Claude Code Desktop to build it, the more I realized, wow, you can pretty much do anything with an agent OS. And you can pretty much build anything that you want into it. So the great thing about this is that we could keep building, building more stuff into it. And the more we used it, the better it got, and the more powerful it became. So like, for example, I had an idea for a lead generation tool, and I built that into our agent OS system. And I basically built like a SaaS tool out of it. So it started as like one idea, and I think I started building it out like probably end of April, something like that. And then from there, just added like more features, more features. Sometimes to test my own creativity, sometimes just out of curiosity, and sometimes because it would save me a lot of time, right? In fact, if you're going to build out an agent OS or if you're going to use ours from the AirProfit boardroom, what I recommend to you is that you look at what time on, and then you start automating from there. So for example, if you look at the latest course that we have inside the classroom, the beginner to expert course here, in the first week, you'll learn how to basically automate anything and build custom workflows around that. And then you can build those into the agent OS, and that makes it ultimately more powerful, more personalized to you. So that's the first place I'd start. And then you can add like another brick, another brick, another brick, right? It's kind of like building a house, but it's AI and it's way more fun and less tiring. And the final question we got here was, I can't tell if this is a real person. This is a real person. This is Julian. But that's pretty crazy that you can't tell that it's a real person. Does that mean we're at the point now where you basically don't know if someone's AI or not? But yeah, this is 100% human. Don't worry about that. So just to recap, we've covered all the questions inside the AirProfit boardroom. I've shown you the agent OS. I've shown you how I think about it and how I build stuff out with it, and also how I build custom workflows. I've answered all your questions today. And also, we've answered all the questions inside the community too. If you want to join the community, if you want me to create personalized videos for you like I've done today, questions, then feel free to join. We've got so many five-star reviews now, which is amazing. For example, all these five-star reviews from awesome people inside the community, which is absolutely amazing. It blows my mind that we've got such a cool group of people who are really positive. If you want to grab the agent OS system itself, it's inside the new daily update section. Grab that over here. And you can grab the zip file to install over here. We show you all of our best trainings and tutorials and that sort of thing. Inside the community, you can ask questions, and I personally answer them every single day as well. So we create personalized tutorials, and then we answer your questions and try and help you as much as we possibly can. So, for example, here, Jeremy was asking about hardware choices, and then we get back to him ASAP. Yeah. And so, for example, like Tobias, he was asking a question as well. We create a personal video for him. Then we noted down all the aspects to it as well, and people absolutely loved it. So that's what it's about. It's just helping people as much as we can. It's our four-week coaching course where you can get help and support in real time. And then inside the map, you can actually meet people who are building out with AI agents like you. So feel free to get all of this inside the AI Profit Boardroom. Link in the comments description or just go to the AIProfitBallroom.com. Thanks so much for watching. I'll see you on the next one. Cheers. Bye-bye.