Agent OS: Automate & Build ANYTHING!

Today we're going to be talking about agentic operating systems, how to use them, how to get the most out of them, which models to use, how to install them to multiple machines, how to build custom workflows and systems inside your agentic operating system so you can get the most out of them. And we're going to be covering some of the latest questions I've had about agentic operating systems so I can help you as much as I can. Let's get into it. So the first question we got was from Jessica and Jessica was asking about giving organizations, staff and team their own desktops. So basically they have a function to carry out which involves taking an exported list of products and then uploading them to like their portal. Now this either needs to be done manually, like if you use Chrome extensions or Claude etc. But with some suppliers, they'd like their agent to log onto the portal, import the sheet, submit the order. And you actually with something like this, what you could do is you could actually have Claude CLI with a custom workflow that submits the products to the supplier, right? So for example, what I usually do if I have a custom workflow that I use every single day, like this one, for example, I'll go into Claude and I'll be like, right, okay, I have this list of products, I need to upload them to a supplier. Here's the API endpoint that I need to use or here's a website I need to upload them for. And I would turn that into a skill that then has a nice UI on the front end so I can just use it in a click of the button. And basically when you do this, for example, Claude can then start building this into your agentic OS for you. So that's the way that I would approach it. The same way, for example, I set up Hermes Muse, Hermes Oracle. I had the idea of voice agent controlled by Hermes. And so I created Hermes Apollo in the exact same way. Just describe what you want, go into Claude, explain it. And then also when you're doing it, make sure that you say, okay, plug this into the agent OS, right? Obviously you need to make sure that you have the product list and the supplier, but also once it's done that, it can easily use that as a skill and then a tab in the agent OS. So it will create a skill first and then a tab in the agent OS, and then you can just access that whenever you want. And also if you, for example, have a team or staff, you could give them access to this via tail scale, right? So they can have access to this via tail scale, which means that they can connect to a machine where you have the agent OS set up. Also, if you want to run it as a schedule task, basically you would also just ask it to have this as a routine. So you could give the endpoint, create the skill, create the tab in the agent OS, and then say, just run this as a schedule on a daily basis. So for example, if we go to Hermes Muse here, this runs on a schedule every single day, right? So it's analyzing content, finding new ideas because it's running on a schedule. And then the same, for example, for Hermes Astros, this is refreshing every single day, right? So if we have a look here, last consulted 11 hours ago, I didn't manually click a button. It just runs on a routine. And you could do that with Claude or with Hermes. Honestly, I'd recommend like trying it with Claude because Claude is a lot smoother, especially if that's what you prefer right now. And it seems like that's what you're using. And then you can just have it running as a scheduled routine inside Claude. So to recap, create the skill, create the tab, then schedule as a routine. Next question we got was from Demetrius. So Demetrius was asking, you know, regarding the agent OS, is it possible to install it on multiple computers? Like for example, if you have a desktop at home and then you want to install the agent OS on a VPS or for example, on another laptop, how would you do it? So you can actually use Tailscale for this. And this works on mobile. So there's a mobile app and you can also connect like multiple machines. It's actually free to try out as well. And then you can basically have your agent OS connected to wherever you want. We have a full tutorial on that inside the classroom. So if you go to the classroom here and then go to new daily updates and you'll find the agent OS and Tailscale tutorial here on exactly how to set up. But it's pretty simple and easy. Like you just create a Tailscale account, install it on your main computer with the agent OS, and then install it on whatever devices you want to connect it to and just toggle Tailscale on whenever you want to use it. We've got another question here from The Secret, which is which model do you use for the agent OS? Right. So they've said, I wanted to ask how you're currently assigning the different models to the agents in the agent OS. So for example, he's using mainly the Olama and they're curious which model they're using for which agent. For example, with FreeClaw Code or OpenClaw, what's the current setup? So for me, I like to test lots of different models. Best model I've tested by far is Claude. The way that I break it down is actually like things changed a lot this week as they always do in AI. So my favorite AI for like orchestrating the whole of the agent OS is Claude. Like Fable 5, amazing model. Then for using an agentic model, DeepSeq V4 Pro, very, very good, very agentic and pretty affordable as well. So you can just plug that straight into Hermes on agentic stuff. And then for example, you got stuff like GLM 5.3 that just dropped and that's kind of matching Claude Fable 5 on the benchmarks. So it's a good option if you just want something that works on coding tasks. That's what it's worked for. But if I had to pick one for one single thing, then it would be Claude Fable 5. And again, you can't really use that unless you use the API inside your agents. So you'd use something like DeepSeq V4 Flash or DeepSeq V4 Pro inside Hermes. And also one thing to bear in mind is like we create separate profiles and just test this stuff out. So for example, when GPT 5.6 came out, we created a new profile for Hermes with 5.6. When QEN 3.8 came out, QEN 3.8 Max, we tested out with Hermes. And so it's pretty cool because then we can switch between different models depending on what we want to do and what sort of things we want to run. So as an example of this, we have like LFM, which is a local model pretty fast for working with Hermes. It was actually designed with Hermes agent harness in the testing, which means like it's pretty fast. Qwen 3.8 Max just dropped. So Victor is asking, you know, has anyone got this running yet? This is actually something we've got a tutorial coming up on. And it's available on Olama as well as LM Studio. So you can run it either way. It's available on MOX, which means like it's better for running on a Mac. It's more smoother and faster. For me personally, I think it's still going to be just slightly too big to run on a Mac Studio. I might test it out, but I just know even like QEN 3.6 on MOX, pretty slow, pretty slow on a Mac Studio. So I think if you've got a DGX Spark, if you've got an RTX, fantastic. If you have like a Mac or a MacBook Pro or something like that, it's probably not going to run that smoothly. And so I wouldn't test it out. But again, it's available on Olama. So you can just test it out with your own setup. And also if you're using LM Studio, it will actually tell you whether it's too big or not to run on your local setup. So you can go into LM Studio. And then from here, this is a free app that you can get. You can search for the models and you can see QEN 27B is here. But you see here that it looks at your setup and it gives you advice on whether it's going to work for you or not. And you can see it says likely too large for me, which I would agree. Just from seeing the size of it, I think it's going to be too big to run on my current setup. Got a question from Niall about AI avatars. Yes. So do we have any training on this? Yes, you can go to the classroom and then you can go to the playbook here and you can grab the AI avatar system that we have. We have a full step-by-step operating procedure. We have a video tutorial on how to use it. If you also want to know how we do like the screen shares and that sort of thing, we have a full system on it here. That's updated based on what my team use. And that's really, really powerful. So if you want to learn that stuff, I'll send you a link. Also, some good news on OpenAI. So this dropped recently, but basically it's been some big price reductions. On OpenAI pricing, which I think they've had to do because if they want to stay competitive, they're competing with, for example, like DeepSeek and that sort of thing. This is really, really good news. One good thing about the community is like people are always sharing what's working for them, sharing their wins, sharing lots of positivity, but also sharing the latest news. So like a lot of people ask me, where do you get your ideas for everything that you talk about? And it usually comes from here because people share, for example, Qwen 3.8 drop-in or for example, OpenAI. And then we test it out and we try it out. But yeah, pretty good news. Also, what I like to do personally is just plug in my existing codex subscription into Hermes. And then you don't even need to pay for the API. You can just use the existing subscription that you have for Chat Cheapity. This is really interesting. And actually, do you know what? I've got a friend that I do tutorials with sometimes. You've probably seen him on the channel, Kazra Dash. And he did exactly the same thing. So for example, in this post, Nye is talking about how basically Boris from Anthropic, the guy who came up with Claude Code, he deleted all his skills and the Claude MD file, and then the system, Claude itself, started performing better. And I've heard that from multiple sources. It's not something I want to do because I've got so many good tools and skills inside there that are very, very, very specific, specifically for SEO and for creating guides like this. And so if I delete those skills, the really, really detailed guidelines I have just won't work so well. But if you don't have like a recurring workflow like that, and you don't have anything specific and you're just using it day-to-day to code or to Vibe Code apps and that sort of thing, then this is a decent idea worth testing. I would definitely keep a backup of the skills and the Claude MD file before you switch anything. But apparently, this does improve performance a lot on Claude. I've heard it from multiple people now. So that's basically it for all the questions. We've covered everything today. For example, making sure Claude performs better, opening eyes, API details, how to use avatars, coin 3.827b, the best models for the agent operating system, how to install it on multiple devices, and then finally, how to give your team and also how to set up loop tasks with the agent OS. If you want to get more training like this and my full agentic operating system, you can get it inside the AI Profit Boardroom. And what I do every single day as well is I just create a video tutorial for you based on the latest questions. So if you have questions, you can post them inside the community. I create a video tutorial for you. And what I do is I post the link to the tutorial and then also break it down into steps quickly in text in case you don't have time to watch the tutorial. But that way, I can create a personalized video for you and help you as much as I can. So feel free to post inside the community. The great thing is there's always people online as well. So you can get help 24-7. Inside the classroom, you get access to all of our best trainings, setups, the agent OS itself, which we're always updating, new trainings, new tutorials that are dropping all the time. Inside the calendar, you can jump a week of coaching calls, ask questions, get help and support in real time. And then inside the map, you can meet people in your local area who are building with AI agents like you. So feel free to get this link in the comment description or just go to the AIProfitBorn.com. Thanks for watching.