Hermes: Agent OS + Obsidian + FREE Apis + WebUI!

Today, I'm going to show you how to build your own AI operating system with Hermes Agent, and it's going to give you superpowers. Imagine having a whole team of AI workers all in one place, working for you day and night. One writes your videos, one edits them, one checks them and makes them perfect all at the same time, all for you. And here's the best part, it doesn't matter what new AI drops next week, your Hermes system just swaps it in and gets even stronger.

You build it once, and it grows with you forever. I'm also going to show you the power of this whole thing, and how you can use it with free APIs and free models and the best ones to use, so that by the end of this video, you're going to have your own 24-7 AI command center that turns your ideas into real things faster than you ever thought was possible. Let's get into it. Hermes with Agent operating systems, I mean, for example, we actually built out this full game, as you can see right here, using the Agent operating system.

And anything new that comes out as well, we can just plug into this system too. So, for example, GLM 5.2 came out over the weekend. No problem, we've already built all sorts of amazing stuff with it, as you can see. And then also, the great thing about it is you can plug in all sorts of cool stuff.

So, for example, N2 with free cloud code came out. We can now control it with our voice and build amazing stuff. We've got an Obsidian memory galaxy over here. So anything that you want to build with an Agent operating system, you can.

And if you saw the situation, for example, with Fusion 5, sorry, Fable 5, and how that got removed, the good thing is, like, if you're building a project, it doesn't matter what models come in and what models come out, because you've got a system that holds it all together, and you can swap everything in and out and just improve from there. So that's the way that I look at this, and I think it's the best way to get the most out of AI right now. And also, the cool thing about this is, like, every time you have a new workflow, you add it into the system. Want to create videos?

Okay, great, we've got a video agent there. Want to automate SEO? No problem, we've got the SEO content pipeline. Want to start building games or creating music?

No problem, we'll add that in, like you can see. And so when new stuff comes out, you can make the most out of it, no matter what model drops or what agent comes out. So let's get into this. The first question we got here was from Scott, and Scott was saying, you know, should you use an agent-operated system, or is it better to use Hermes Desktop?

And for me personally, I would say it's much better using agent-operated systems, because, for example, if you pull up Hermes Desktop, like so, let's open up Hermes Desktop. The problem with Hermes Desktop is that, number one, it's only for Hermes, so if you have other agents working, for example, like Claude, how are you going to operate them inside Hermes Desktop? You're not. Whereas, for example, we can have a group chat with all of our agents in one place, with Claude and Hermes and everything else in one place right there.

We can also, for example, if a new model comes out, like GLM 5.2, we can have a separate chat here for GLM 5.2. KimmyK2.7 comes out, no problem, let's chat to it right here with a separate profile. And so, I also like that. And the other thing I would say here is, like, you can build more custom stuff that's more visual.

So, if you look at Hermes Desktop, like the UI is super simple. You can't really preview and see the amazing stuff that you've built with it. It's very hard to come back to as well. Whereas, for example, if you look at Hermes Jarvis that we've built over here, well, that's ready and waiting whenever we need it.

We've got, Kimmy as well, with the CLI over here. You can't really do that with Hermes Desktop. And so, the biggest differences I would say here is, like, number one, you can manage more agents with an agent operating system. Number two, if anything new comes out, you can build that in.

And number three, you can make it much more visual. Hermes Desktop is very, very simple. And it's not as fun to use Hermes Desktop as it is to use an agent operating system that you've built yourself that's optimized in exactly how you want it. By the way, if you want to ask questions like this and get me to answer them inside a video tutorial like you're seeing today, then just join the AI Profit Board and you can post inside the community and then I'll answer your questions about agent operating systems or whatever you need help with.

This is pretty cool. So, Jason is building out his own sort of avatar for the agent operating system. So, basically what he's done is he's got our agent operating system zip file and then he's improving it and making it better and building on it. For me personally, I mean, this is one of the great things about this is you can add your own images and you can customize it how you want.

I mean, look how cool that is. I would say personally, this one is my favorite. This one is my favorite, but they all have a really cool vibe. This is a great question by Wes.

So, Wes was saying, you know, he sees a lot of chat about agent operating systems with different integrations with different LLMs. So, how do you manage like the costs and the resources involved with that? So, the way that I would look at this is, for example, if you look at how we use Kimi Code and GLM 5.2, it doesn't really require that much because we've got the setup with the coding plan. And so, everything that we build inside here is not capped.

It's not charged by API. So, my first recommendation would be using the coding plan, for example, like Minimax M3, Kimi, K217, GLM 5.2. They all have coding plans, which means that you don't get charged for the API. You just have the coding plan ready to go.

The other option that you have is you can use free APIs on OpenRouter and these are coming out all the time. So, for example, here, look at this. This is looking great. All right.

So, if you come back to OpenRouter and we type in NEX2, or if you type in free in OpenRouter, you actually see loads of different APIs like this and our alpha. There are loads of free APIs on OpenRouter that you can use and plug into the system, which helps you too. And then the third option is you can use a plugin. So, for example, there's open source projects like Headroom and with Headroom, you can reduce the amount of tokens you use with any LLM.

And finally, for example, if you're using Hermes or any other AI agent like that, then you can log in with OAuth using, for example, Kimi's coding plan or Minimax's coding plan. And that means, again, you can just use the coding plan that you already have instead of having to worry about APIs. Good question. This is a separate agent operating system that Benjamin's started building out.

So if you look at this, he's got his memory galaxy just like we've built. So what he's done again is like, this is a great thing about having an agent operating system and what we share inside the AI Profit Boarding, is if you go to, for example, the memory section here, he's got that style from the zip file that we share inside the AI Profit Boarding for the agent OS, and then he customized it exactly how he wants, right? He's made it his own. So if you look at this, for example, he's got Claude, he's got Gemini and everything else plugged in, but the UI has a total different feel to it, a different vibe to it.

And it's basically his own style. So this looks absolutely amazing and a great example of what you can do and what's possible when you're building out an agent operating system. Really like the fact that you've made it your own and also tweaked it exactly how you want it. Looks amazing.

So you can see great examples of how people are just building out awesome stuff like this. Well done to Benjamin. But it's a good example of like, you know, anyone can do this. This is an interesting one, like how do you organize your Obsidian vault, especially if you're dealing with different file types like you can see here.

So the way that you could do this, one way you could approach this is you could get your agents to look at those documents and then take notes and put them inside your Obsidian folder. And obviously if they're inside Obsidian locally, then you can access them offline whenever you need them. There's some pretty cool skills as well, like you can see here for this exact job. Just check the skill MD files before you install anything or you could ask Claude to use that as an example, but then build your own skill that can help you organize your Obsidian better.

By the way, if anyone's watching this and they're like, what is Obsidian? What is that? So basically Obsidian is like a second brain for your AI agents. So my agents take notes from every single conversation and everything that they do with me.

And then what they do from there is they plug that into this Obsidian vault that is basically a second brain of everything that I've ever done or organized. And the great thing about this is if you can store it as markdown files, so it's accessible local, you can also get an MCP to plug it into your AI agents, if they're, you know, on a VPS or something like that. Then once you've done that, you can plug that into your agent operating system and your whole system revolves around this. So what that means essentially is like, if we are creating new things, like you can see here or chatting to our agents, that gets logged automatically into our Obsidian vault so that our AI agents have context in all the previous conversations.

Even if, for example, using Claude or Hermes, they share the same brain and they share the same memories, which means that we can easily move from one project to another and organize everything beautifully, which is a fantastic way to make things easier when it comes to context. And this way as well, you don't need to keep, like, giving more information to agents or rebriefing them, which I think people probably spend like maybe 15 to 20% of their time doing. Also, the cool thing is, well, like, if it's got context, then you don't need to go back and forth with it as much, which means that you save tokens as well. This is a great question from Farah, who needs help building out like, you know, video with AI agents.

So my workflow for this is, we actually give our agents access to hyperframes as a skill. So Hermes can use hyperframes as a skill, and then it can create awesome videos. So if I show you an example here, this was with KimiK2.7 today, and we've also done it with GLM 5.2. If we play this back, you can see that we've got a fully edited video.

I didn't touch any of this, but it scripted it. It created the AI video itself. It generated the voice as well, and it edited it properly as well, even with the camera angles and everything like that. So the way that you can do that is, you can go to your AI agents, give them access to hyperframes, which is a skill as you can see here.

It's a free skill for creating videos. And from there, what you can actually do is connect it to Haygen as well, if you need avatar videos. Additionally, you can give this to Claude or Hermes. It can work with any AI agent.

And the great thing about this as well is that you can have multiple agents build in it. So for example, for us, if you check out this Kanban board here, we had a team of different agent profiles. So we've got like a video writer, video editor, a video judge that makes sure it's actually good and iterates it until it's finally completed, and scores it as well. And so basically they can keep going and iterating on the video until it's finally good.

Then we have one agent dedicated for each part of the process. So one for editing, one for writing, one for judging if it's actually good. And they just keep iterating until the judge actually says, wow, this is amazing. Example in action below.

So that's a really powerful use case of an agentic operating system because you've got a Kanban board here, you've got all your agents plugged in. So with different profiles and everything. And then also we have the video agent down here as well. So again, like if you want to create content or you want help with social media, et cetera, then you can build your agent operating system like we've got here around that whole thing.

And that's one of the best things about this whole system. So Jason is asking like to NA10 or not to use NA10. Which one would you use? And should you carry on using NA10 or should you use, for example, an agent operating system?

So for me personally, if I'm using NA10, like it's a lot more technical. I have to go inside it, I have to edit things. It's going to take hours to create like an amazing workflow like this. Whereas I can build it in minutes with, for example, Claude building into my agent operating system.

So for me personally, because of my experience with AI, I would pick an agent operating system every single time. And again, it's just much easier to visualize everything. Whereas for example, if you were trying to build this whole system with NA10, it would get very messy and it would break a lot. So if you're comfortable with AI and you already understand the foundations, I would just build an agent operating system.

I wouldn't use NA10. Because an agent operating system is much more complex, but easier to navigate, especially if you've got Claude or Hermes building it out for you. Whereas for example, if you are using NA10, things can break a lot. It's quite technical.

It takes hours to set up. It's a lot longer to set up. And also it doesn't link together as nicely of a beautiful UI. This is a great question by Yoche, which is like, you know, when you've got a local AI agent and it can do stuff, how do you limit it?

So one of the first things that I avoid is giving access to my main email address. I actually give it its own inbox and that way it can't do anything crazy in my main inbox. And also there are guardrails built into, for example, Hermes and Claude that stop it doing things it shouldn't do. But I think you can always give it more rules in terms of how to operate and also the permissions it uses so that it doesn't go off and do anything wild.

I think that's a good way to train it, is just set up more rules as you're thinking more stuff that could help it be guided in the right way. This is a great question, right, where Blake is asking, how do you keep Hermes awesome but also avoid using a lot of tokens? So there's a few things I would do here. Number one, you can use free APIs and there's a lot of good ones, for example, like Entoo just came out.

Number two is you can actually give it the open source skill, Headroom, and that reduces the amount of tokens it uses each time. Additionally, you can use the free models via Noose Portal if you log in. So for example, Step 3.7 Flash and Nemetron 3 Ultra from NVIDIA are available for free if you log in with OAuth on Noose Portal and Hermes. If you have a good setup, for example, like an NVIDIA Spark, then of course you could run local models as well and that would help a lot too.

So if you're watching this and you're like, okay, well, how do I get those free models? So what you can do is you can go into, there's a couple of options, but if you go into, for example, Hermes here, and then you go to Manage, then you go to Models, you can change the model inside your dashboard. And then you can, for example, select Noose Portal and from there, you would select the free models. So if you select Noose Portal here, you can see that we can set up Nemetron 3 Ultra and Step 3.7 Flash for free.

So this is a question about Jarvis, Hermes Jarvis that we've got here plugged into the system. So when it comes to Hermes Jarvis, like it's basically, it's a voice activated agent, which means that it can be a little bit buggy sometimes, but it usually works pretty well. So if we test out an example here, Jarvis, can you just open up juliangoldy.com for me? And you can see it's thinking now.

So it's thinking about it and now it will open up the website as you can see. So it actually works and then you can see the full transcript here and that sort of thing. But if you're having issues with it, usually what I do is I go directly into Claude and I ask Claude, like, hey, how do you fix this? Or here's the issue that I'm having.

Can you just improve that part of my agent OS? And I just keep testing it back and forth until it's fixed. Usually it's good as well if you can add a screenshot, but as well, Claude can see like the recent logs inside your agent OS, or if you're using Hermes, you can do the same thing. And then also you just want to double check what models you're using for it and make sure you have the latest version because we update it daily and any sort of bugs that I find, I usually fix.

So for example, if we're trying to get the latest version of the agent OS, we actually upload a new version here once we've added all the new daily updates inside this section. So you can see the agent OS, the video tutorial on how to set up, the last update date, a full installation guide, and then also the resources on how to use it. So this is a good question from Sam, which is, you know, what changed with the agent OS yesterday? And then also, can you run agent OS on a VPS?

So we actually have a really good tutorial here. This tutorial shows you how to set up the agent OS on a VPS. So if we open this up, you can see that John actually created a full guide on how to set up and how it works with previews and everything else, which is pretty cool. For the changes in June the 13th, what I'll actually start doing is adding a new change log inside the whole setup.

So you can see what changed day to day. On June the 13th, what we actually added that was new was Kimi K2.7 and the new CLI inside there. And then also we set up a Kanban system where you can basically have a video agent iterating on everything that's been improved. We've got a couple of examples below.

So Harry was asking like how to set up an agent OS. So you can get the full zip file inside the AR Profit Boarding for the whole setup. But it sounds like you've already got that, Harry. So for setting all of this up, like for example, you mentioned you want to set up lead gen, outreach, meeting notes, et cetera.

I would just focus on one thing at a time and depending on your skillset, obviously that will change how long it takes to set up each process. But I would just focus on one thing at a time. So for example, for lead generation, what you can do is start building that in now. Ask Claude like, hey, I want to set this up.

Here's the idea I have for lead generation. Here's how I think it should work. And then Claude will actually tell you exactly how to set it up inside the agent OS and it will build it in for you so you can tweak it exactly how you want. So for example, anytime I have a new idea, because I've done this so many times, I can get it built within a couple of hours.

For you, it might take a bit longer because it's the first time, but it's a skillset you build when it comes to improving your agent operating systems. And then any APIs that you need for lead gen or anything like that, Claude will actually tell you which ones to get and then you can plug it back into the system. But again, it sounds like you have a lot of stuff you want to set up, just focus on one thing at a time, simplify it. Otherwise you'll feel overwhelmed and it will be a bit of a struggle.

Whereas if you simplify it, you set up maybe one thing each week, that's going to be way faster and way less stressful. So this was interesting about Fable 5. I already did a tutorial about it and announcement about it yesterday. But basically the great thing about having an agent operating system, this is something you should pay attention to if you don't already have the setup, is that it doesn't matter what agents come out.

It doesn't matter what models get removed. This is very, very flexible as a system. And that's a great thing about having a system like this. It's like, you know, if Fable 5, you can't use it anymore, no problem.

Let's add in KimiKator 7. Let's add in GLM 5.2. Let's add in Gemini with the cloud managed agents, right? That's all sorts of stuff we've done in the last 24 hours.

So as you use this more and more, you'll see more ways to improve it and evolve it. And it just becomes a system where it doesn't matter about the model, doesn't matter about the agent. It's a system that you've built and you've customized yourself. And that's a beautiful thing about this.

But I'd expect Fable 5 to be back soon. I think this will just be a Terran Prairie situation. Hopefully, you never know. This is a good question.

So Nicholas is asking like, what's your take on Kimi code? You know, is it useful? Is it not, et cetera? And so I really like it so far.

It's been pretty good. Obviously the coding plan is pretty decent as well. I've not run out so far. I'll show you some examples of what we've built here.

So for example, this was a fully edited video with an AI avatar. Over here, we created this game. This was okay. I think it could be a lot better in terms of the graphics and stuff.

This was pretty cool. This was another interesting one. And we created this game. One of the most impressive things that we created was this.

So this was like a 3D game that's kind of fun to play as you can see. So I think it's pretty good. Is it as good as Claude Opus 4.8 or Fable 5? No, of course not.

But does it do the job and can you build a lot of amazing stuff? I mean, do most people need the full power of Fable 5? No. So overall, I really liked KimiK217.

I think it's great on the coding plan. I think you can build a lot of stuff with it as well. And we've also just plugged it into the agent operating system too. And we've got a full tutorial on that as well.

So Amanda built this out as well. Let's take a look at this. Amanda says the mothership has landed. So she's wired Hermes agent into an agent operating system.

It's actually working. You know, she's got the Hermes tab with the chat sessions, workspace, et cetera, which is pretty amazing. Running on this, wow, wow. So running it with Olama and Quant 3, which is pretty impressive.

Let's have a look and see what we've got. I mean, it looks super nice. Look at the way that's organized. That looks nicer than my UI.

I should step up my UI on that. That looks really cool. But you can see the vibe and the feel of it. That's pretty awesome.

I'm actually inspired to improve my own UI now. Thanks for sharing. Keep us posted on new updates. By the way, you might be watching this and thinking like, that's great for Amanda or that's great for them, but I can't build this myself.

I've seen so many people who have never used AI before get absolutely insane results with this. So if you're thinking about using it, but you're like, I'm not sure if I should do this, blah, blah, blah, I would say it's great. It's really, really good. And you know, if you're non-technical, that's okay.

I've seen so many people winning with AI this year. You know, we've got 182 pages of testimonials and wins from the AI Profit Boardroom. Like you can see like Amanda, you've got Ken built out his, Nick, Ethan, 61, not a coder, Steelbelt built out his agent operating system, Eric as well. There's just so many wins and so many people building amazing stuff that, yeah, for sure, like if they can all do it and I can do it and I'm not a coder too, then you can do it too, right?

And it's just so much fun. It will make using AI and operating systems in general, way more fun, way faster if you build this. This is another spin on Hermes Jarvis. And basically what Benjamin has done here is created his own Logos Oracle that he can speak to.

So if he says, okay, you know, what is the future of artificial intelligence? It will actually answer him, as you can see right here. And the UI is super nice as well. So that's just another way of showing you how you can tweak this and how you can make it better.

And this is a great thing about having a community as well. It's just like we can all share and we can learn or we can grow together and just help each other. You know, you can see all the positivity going on here. So people just sharing wins and creating amazing stuff.

This is pretty cool. So Gene was talking about this, which is another open source project called Mark It Down. And basically what this does is it turns files into Markdown and then you can use that as a useful way to feed and input the LLM. So that's another way to reduce tokens as well.

So thanks so much for watching. You've seen what you can do with agent operating systems. You've seen how they work. You can see how powerful they are as well and what you can build with this stuff.

It's absolutely amazing. If you haven't built out your own, definitely recommend it. If you want to get minus inside the AR Profitablearium, you can see all this cool stuff that we've built with it. And you know how much fun this is to build with as well.

The other cool thing I would say here is once you get your head around it, like you'll feel unstoppable. Like the thing that motivates me every single day with agent operating systems is like, I can build something new into it. I can create something amazing today. I can make Hermes Jarvis even better.

I can plug in Kimi K 2.7 and build out like these insane games that we've got here and all this cool stuff, right? And so that's the best part about all of this is like, you get a system that you make your own, you improve every day. And every time you have a great idea, you go from like idea to implementation ASAP. We even have this, for example, app builder, where you can plug in an idea and then build it.

And all you do is have to reject or approve the plan. And then it goes off and build stuff like you can see. So there's no limit to that. I mean, this is something that I've only been building out for a few weeks and it's just getting better and better and improving all the time.

So definitely recommend it. If you want to get my setup, you can get that inside the AR Profitablearium. Link in the comments description or go to the arprofitablearium.com. And this is a great place where like you've seen today, I answer the questions personally, in a video tutorial like you've seen.

You can also get the AgentOS system inside this section with the video tutorial. You can see it's last updated date. So you can see it gets updated with new cool stuff every day. You can grab the zip file with the installation guide.

We've also got new cool stuff. So new tutorials and video tutorials coming out daily too. Inside the community, you can ask questions and you get help and support. There's always people online 24 seven.

And that's a great thing about this as well. And like I said, I personally answer this stuff. Everyone inside the community helps each other. It's a positive place to learn and grow on this journey together, where we're all learning new things, right?

We're all progressing and we're all improving and just using this opportunity and time to make most of things. Inside here, you also get all of my best trainings like you can see. So for example, if you're a complete beginner, you can go from beginner to expert in just six weeks and also have new daily updates if you prefer the more advanced stuff like you can see. You also get four weekly coaching calls where you can jump on a call, ask questions in real time, share your screen.

And inside the map, you can connect with people in your local area who are using AI agents just like you. So you can meet people in your local city who are using agents and operating systems like you're seeing today. And I've shown you so many examples of people building their own. I know if you're watching this and maybe you're new to AI or you're thinking about setting this up but you haven't, this is a time, right?

There's never been a better time to create this stuff because AI is at a point now where it's so much fun. So hope to see you inside the AI Profitable Audium. Thanks, cheers, bye-bye.