Hermes Agent OS: Build & Automate ANYTHING!

Ajouté : 2026-05-23

[00:00:00]Today, we're going to be running through some of the newest Q&As I've had about Hermes agent operating system, how to use it, how to use Hermes agent, how to use Open Claw, how to create teams of agents. We've got a bunch of great questions that I'm going to run through today, and we'll be running through exactly how to handle all of these different challenges, all right? So, let's get straight into this. The first one we had was from Azeem, right? So, Azeem actually posted that he's looking for help on how to basically come up with video generation alternatives to HeyGen, right? So, what alternatives to HeyGen could you plug directly into your AI agents? And he said, you know, I'm looking at HeyGen, um but if I'm on the free plan, like you don't get much on it. What are some good other options, right? My answer for this is that you actually don't need HeyGen for daily video content, right? What you can actually do is you can run video generation locally for free. Um the free path is using HyperFrames, right? So, HyperFrames is an open-source AI agent that runs locally. You can build animated videos for you, and you can easily plug it into your AI agents. So, whether you're using Open Claw or, you know, for example, Claude or whatever you want, you can plug this into a video generator. Let me show you an example of that. So, we have a full video tutorial on it over here.

[00:01:30]And basically, you can create videos for free with that. So, you can use HyperFrames, and then you can plug it into Hermes agent. Hermes agent can actually trigger and build the videos for you, or you can use any other agent that you already run. So, that could be Claude, Open Claw, whichever your stack is on. And then, the output will go to your local assets, right? So, let me show you an example of that in practice.

[00:01:50]If we have a look at Hermes over here, we actually installed HyperFrames directly into it. Um with HyperFrames, I'm going to show you an example here.

[00:01:59]We actually created a video animation and then it saves it locally, right? So, we said, you know, use this to create a video and then we got this right here, right? And and that's the example video we generate. It's just a quick 5-minute video, but you see how you can generate videos for free using HyperFrames.

[00:02:16]Another good option is Remotion as well.

[00:02:20]So, a great question there.

[00:02:45]Next up, let's have a look what else you got here on the questions. So, you have a good question from Michael.

[00:02:51]Um, Michael posted you know, he's looking for a little help with setting up local AI, right? Um, basically, he said, you know, he's upgrading from a Mac M2 to a Mac M5 Pro with 48 GB. You know, four little sub-questions here. So, you know, is the new Mac enough for Agentic OS and Hermes Claude code and local model? Is Open Claude required for the stack? Should you sandbox each agent? And which Claude model do I use to build Agent OS? And does it have to be Claude? So, lots of good questions here. Let's go through them in order. And by the way, if you're not familiar with Agentic OS, this is the full system that we have.

[00:03:31]You can build your own, of course, but if you want to get mine, it's inside the AI Profit Room.

[00:03:35]You can see here that we've built out, for example, all of our agents working together with Open Claude, Claude, Hermes, Gemini, Anti-Gravity, Free Claude code. And then we have all these cool tools built in here, too. So, first question, let's go through these in order. Is 48 GB enough? Yes, comfortably, right? The Agent operating system, Hermes, Claude code, they would all fit. Now, do I recommend using local models in general? Not particularly, because I don't think they're as powerful as something like Claude or, for example, even, you know, free API like OpenRouter with our alpha.

[00:04:09]But, if you're insistent on using local models, then, yeah, for sure.

[00:04:14]Is Open Claude required? No. You don't You don't need to plug in Open Claude into your agent operating system. Agent OS works fine with like Hermes and Claude code. Open Claude is a powerful addition for specific automations, but it's not required to start. Should you sandbox each agent? That's a great question. So, it's good practice for production work, but it's not required to get started, right?

[00:04:38]So, it's up to you.

[00:04:40]Personally, do I? No. Which Claude model to use? So, you know, just using Claude code is great. You can actually plug in a free API into, for example, Claude code. We've got free Claude code here.

[00:04:53]It's running with our alpha. So, that's a free API if you don't have a subscription to Claude normally. And then, you can use this to build out your agent OS, right? Agent OS actually runs locally as a setup, but it is powered by APIs, right?

[00:05:08]Does it have to be Claude? No. Free Claude code uses our alpha and OpenRouter. That's also free. And you get the same harness, but it's, you know, a different API.

[00:05:17]And on the GUI thing, I think you mentioned it as well. So, you mentioned that you don't like GUIs. Um and you prefer a GUI. So, if you don't like the terminal, that's fine. Like, agent operating system is a dashboard.

[00:05:30]So, you can set it in terminal terminal once, and then everything happens through buttons and panels after that.

[00:05:34]You won't be living in terminal.

[00:06:01]>> So what I like to do is just like answer everything with a video tutorial and then also give you like the bullet points in case you don't have time to watch the tutorial on the video.

[00:06:16]Mohammad says, "What about the new update of Anti-Gravity 2.0?" So we actually plug that in to Anti-Gravity.

[00:06:21]The cool thing is with the new update from Anti-Gravity, it has a CLI built in and so with that CLI, you can basically um you you can plug that into your Agent Operating System for free, which is pretty nice.

[00:06:34]Joe has a great question or a a great win here, I think actually. So Joe said, "Today we set up the Hermes Agent Team, with each agent having a clear role." So he's got an orchestrator, a workspace, a content writer, the thumbnail creator, uh review editor. Took a month of testing, breaking things, fixing things, but the first article is published live.

[00:06:56]So congrats to Joe." You know, Joe has just shipped what most people get stuck planning forever, which is a real working multi-agent team using the Agent Operating System and you know, it's a repeatable workflow. So Here's what I would recommend, you know, based on what's working for Joe here as you can see. And there's a lot of cool like really cool developments people in here sharing lots of value, which I love as well. You can see he shared the full setup right there.

[00:07:20]So basically to build your agent team, right? You want to give each agent one specific job, right? Not marketing assistant, um sorry, you know, not marketing assistant but thumbnail creator for example.

[00:07:34]That's a a very narrow job. That's better. You can name them, give them names, makes the system easier to think about and easier to debug. Give each agent skills and memory so they're not random chat windows starting from zero.

[00:07:45]Build a work proof layer so you can see what they're producing instead of just watching icons move and then test end to end on real content, right? But huge shout out to to Joe. He's he's crushed it. So that's awesome. And you can see for example, if we go to the Kanban section here, we can manage our team of agents and we have a swarm of agents here as you can see. Like some for SEO, some sort of like a persona of me and we can just plug these in, which is really Let's have a look what else he got here.

[00:08:46]So Georgia was saying, you know, how do you go about identifying the current digital trends on like Instagram or LinkedIn or TikTok? You know, are there any tools that you currently use? So here's I my answer, right? I have a day I have a daily ideation process that I use.

[00:09:04]And so essentially there there's a few different things you can use for finding new ideas. And this is something I'm constantly working on, right? So number one is you can analyze daily, you know, from the content you've posted this week for example, you can update the latest views and just see, okay, what what's getting the most attention this week and then how can I create more of that?

[00:09:23]Number two is like use a custom code project for more ideas based on the trainings and the data that you already have so you can build them.

[00:09:30]Analyze like the biggest pages on Twitter who cover new stuff in your space. Then analyze like your previous best performing content, right? What made them hit? Why did they hit? What's trending right now? Additionally, search socials for like related terms in your industry. So, for me, for example, it would be like open core and Hermes right now. That's what's trending. Check the latest open source projects. Usually a really good early signal of what's trending. And you want to check trending products on GitHub as well. And then look at what worked recently and remake it with a fresher angle, right? Check trending topics on Twitter.

[00:10:05]Pick a list of your competitors as well in their social profiles. Check Google Trends. And then also journal each day on like what went well and what to improve. That will give you more clarity on how to get better ideas.

[00:10:41]Um here's what's interesting. Like the actual edge here is like most people copy what's trending, but the journaling is really what makes this compound because every week you have a record of what works specifically for your brand, which is what matters most, right?

[00:10:53]That's the most important part.

[00:10:54]Let's see what else we got here. So, Surf a Lot says, "Does free Claude Code have all the same capabilities as regular Claude Code?" Here's what I would say. So, yes, anything you can do with Claude Code, you can do with free Claude Code, right? And here's why they're equivalent. So, has the same harness, which is the Claude Code CLI.

[00:11:13]That's identical in both. You have the same commands. So, every flag, every workflow, every plugin that works. You have a different API under the hood. So, for example, free Claude Code that we have inside the agent operating system here is working with our alpha on open router instead bit Anthropic's API. Why?

[00:11:31]Because it's free.

[00:11:33]And then, you know, it's connected uh locally. So, that's the whole trick, right? Same CLI, same UI, but just a free model.

[00:11:43]That is a great question.

[00:12:11]Andreas was asking He said He said uh I'd like to set up Agent OS at home on a Mac mini with the M1 chip.

[00:12:21]Would that be sufficient? Yeah, I I think that would run, you know? And you can always make this more lightweight as well later. So, you know, for the dashboard, connecting to cloud APIs, um yes, it would work. Would it work if you also want to run local LLMs alongside it? No. But, you don't need to do that.

[00:12:39]I don't do that, right? So, the Mac mini M1 with 16 GB can run the Agent operating system dashboard, like you can see.

[00:12:49]It can run Claude code, Hermes, and free Claude code, all using cloud APIs. It can run a small local LLM alongside, but I think that'd be tight, and the RAM would be the bottleneck. So, I probably won't recommend that, but it could run multiple agents calling the cloud, right? That'd be pretty comfortable. So, my recommendation is like start with the Mac mini M1, route through free our alpha for the models, build the dashboard, don't run local LLMs until you actually need them. And bear in mind, like 99% of people don't.

[00:13:19]Great question, Andreas.

[00:13:48]>> Next up, we got Mark. Shout out to Mark.

[00:13:51]And he was saying basically the Gemini 3.5 flash, anti-gravity and Agent OS training is an absolute game-changer for his SEO agency. So, he said if you haven't seen it yet, suggest you get on it without delay. And that's, you know, that is absolutely awesome, right? Like wins like this are the whole reason the boardroom exists. Real agency owners using Agent OS to help clients, that is absolutely amazing. If you're not sure which section he's talking about, you can get the full tutorial over here.

[00:14:22]You can use this for your SEO agency, so shout out to Mark, what a legend. And Mark stack is the exact one we cover in the latest training, right? Three pieces working together. So, the SEO agency stack that we use for the Agent operating system, you can see we have a SEO section over here. So, you have Gemini 3.5 flash, then you got anti-gravity 2.0, and you have Agent operating system, right? Which is a dashboard that connects them, holds the memory, and tracks every deployed website. So, together it really changes the entire way you would run an SEO agency cuz you don't need to write any content anymore. You can deploy content faster, and also you get clients more results better, right? And and faster.

[00:14:59]So, shout out to him.

[00:15:23]And if you have any questions whilst you're you know, whilst I'm answering these ones, feel free to leave them and I'll make sure I answer your your questions.

[00:15:32]So, feel free to post them in the chat and I'll make sure I get round to answering them.

[00:15:36]So, Carlos says, "I just joined the community. I'm really excited. What caught my eye was information about Hermes OS.

[00:15:45]Where exactly is it?" So, here's the full module, right? You can get it over here. This is where it is. And that has the zip file for Agent OS. It has the prompts and the instructions, etc. So, you can get that with the readme instructions inside this module.

[00:16:10]And you know, when you're new to this, when you're new to the Air Traffic Bot room, like the classroom tab is where everything lives, right? So, you can bookmark this one specifically and then, you know, most members will come back to it weekly.

[00:16:41]We got an awesome awesome post here which says, "You're the goat. If you guys haven't joined the Air Traffic Bot room, join immediately."

[00:16:49]Thank you so much, man. Thanks for posting that. I appreciate that.

[00:16:52]That is great. So, Tom has posted as well.

[00:16:58]After setting a goal, I finally have a win. Had a small issue to solve with a button that didn't work. Took longer to find that bug than to complete the build.

[00:17:05]It's V1, but it does what I need to right now. So, he Tom actually built his Agent Operating System.

[00:17:10]And he he's basically created it, you know? So, Tom did the most important thing, which is create version one for the agent operating system, right? And so, if you look at Tom, basically he started because he set a concrete goal, then he built the basic functions first, he debugged as he went, and then he created, you know, version one without perfecting it. You don't need to perfect it. And now, Tom has a real agent operating system that he can iterate on every single week, and that's the entire game. So, shout out to Tom. Well done, man.

[00:17:46]Tony says, "Just started using OpenClaw and Hermes to run a fully automated system.

[00:17:54]What's the best way to build an audience?" Right, this is a great question. So, I run these systems daily, and here's what actually moves the needle for me, right?

[00:18:05]I'm going to break this whole this whole section down for you.

[00:18:13]So, you get the most out of it.

[00:18:31]You know, for me, the brand building systems I use, right?

[00:18:34]So, for example, one of the best things about AI is that you can scale the content. So, you can create from one single idea a video, a blog post, an infographic, social media post. So, for example, we'll post something on LinkedIn, and then we'll also post it on Twitter, and we'll just multiply the content that we create so that we reach more people across more platforms with the same idea, right? Next up, the daily ideation process that I just mentioned earlier in this video, you want to run that every single morning. Um the agent operating system is always a really powerful way to build a content factory, right? And you can also journal inside there on what's working, and you can use Hermes agent to implement this stuff, right?

[00:19:14]So, for example, for actually creating and deploying blog posts, we'll use the SEO section over here. For generating videos and images for social media, we can use this section, right? If we want to generate podcasts, we can do that with NotebookLM. And so, you can create a lot of content that's really powerful um across multiple different places. And then also, community first publishing, right? It's a show where your audience lives, right?

[00:19:37]Don't wait for them to find out.

[00:19:39]The more you can reach your audience, the better you're going to do here.

[00:19:42]I would also document any systems that you'll be building publicly, cuz every workflow you build becomes content, too, right? Like, if you're building a new process, share it, record it, show people, right? And then, the thing I would say here is like, you're already further than most, right?

[00:19:58]It's a If you're running across six platforms already, um that's great distribution. Now, the work is just making each one feel less like an automated feed and more like like you, right? More authentic to you.

[00:20:09]So, shout out to you.

[00:20:21]We have another question from Sirflo.

[00:20:23]He said, "So close to what I want.

[00:20:25]I need free code code in the mission control panel of AgentOS, plus Obsidian, so my team of agents never forget what we're working on. I'm not technical, need something I can just paste to Manas, right?" So, if you actually want the zip files, we have them inside this section here, right? So, you can actually get the full setup that I've used for AgentOS with free code code inside the zip file here, right? And you can just give that to your AI agents to build out for you. So, if you're using Manas, it can set up that with Manas right here.

[00:21:00]So, you know, here's what I would say, you know, you're closer than you think.

[00:21:05]So, number one, here's the instructions, right? You can give Manifold the GitHub URL.

[00:21:14]You can tell Manifold to install it.

[00:21:16]Tell Manifold to set up the API from OpenRouter, plug it into Agent OS, and then wire it into Obsidian. And that's a simple process, right? Simple five-step process. Do one thing at a time. Don't try and rush it all or try and do it all at once cuz you'll overwhelm yourself, but just do a little bit here, a little bit tomorrow, you know, and from there you'll you'll create something awesome.

[00:21:47]You know, and one more thing, I know, for example, Serf Loss, you know, was asking about using Claude with Claude design. Claude itself is genuinely great at design, right? So, for example, if you look at this guy today, we've used Claude to help us create that. And so, you can ask Claude inside of Claude Code to design new panels for you, and it builds them directly. It's really really really good at that.

[00:22:06]Let's have a look at the next one.

[00:22:09]So, Thomas was saying, "I want to build a web app where you enter URL, that runs a AI search engine aware SEO audit, then connects to WordPress to fix the errors automatically with auto backup." You know, um and he's looking for collaborators, which is great. Like, this perfect place. You can see, for example, Justin is connecting with Thomas. So, it's an awesome place to find like partnerships and great people to work with.

[00:22:31]I think it's a solid idea. The auto fix piece is really good. Most SEO audit tools stop at here's a list rather than actually implementing for you. So, here's how I would look at it, and it works in four stages, right?

[00:22:46]Basically, first of all, you want to get the Ahrefs API, which is like the best signal for keyword and page-level data.

[00:22:53]Then you can create a Claude dashboard where the user enters WordPress login details and API key. You can pair that with Ahrefs and WordPress API, which means this combo gives you everything you need to audit and also fix that stuff and then you can auto deploy the fixes, right? So, Cloud can log into WordPress through the API and push new updates with a snapshot. That's the way that I would do it.

[00:23:31]And we have a question from Vinay here.

[00:23:34]Shout out to Vinay. Who says, you know, what's been what is like one thing that the clients are really looking for, right? Is it for example GEOs, is it setting up AI agents, is it content systems, is it lead gen? My personal answer is that lead generation is the thing that people really care about, right?

[00:23:56]And the reason is simple like clients want to see a clear output for whatever they hire you for, right? They don't really care about the tool, they don't care about the tech stack, they don't care whether you call it GEO or AI content, they only care about one thing, right? Which is the number of leads, the number of qualified leads. That's pretty much it. So, you really want to focus on something that goes into that. Now, what could go into that? It could be like AI avatar videos, it could be AI generated content, it could be SEO content swarms, could be cold outreach, right? But, the main point is you want to focus on the end result, which is lead generation.

[00:24:51]So, I think we've answered every question that we've had today, uh which is pretty awesome.

[00:24:57]You've seen all the wins as well. I mean, we've got such a great community of people, like 161 pages of wins of of people learning and growing with this stuff, which is pretty awesome, as you can see right here. Um the main thing I would say is like if you want your questions answered like this daily, join the AI Profit Boardroom community. I do video tutorials every single day. I mean, you could do this alone, like this guy over here, right? Um but you shouldn't, right? But like learning AI automation should be fun. It should be part of a journey. You should share that journey with other people. You should have people who help you move faster, right?

[00:25:33]And the thing that I would say here is like you can do it solo, but you're going to get stuck fast, right? Let's say for example, you hit a wall. You know, you're struggling with a problem with AI automation. Well, you might Google it, you read like three outdated blog posts, and you watch a video from 6 months ago about it, you try all the steps, they don't work cuz the tool is updated already. You might post something on Reddit, get like one helpful reply in 2 days, then you give up, switch tools, hit the same wall again, right? And repeat for every single problem. That's basically what 99% of people who are stuck with AI automation are doing.

[00:26:07]With this way, you know, together, we're working together, we're helping each other create. So, if you hit a wall, you can drop the question inside the AI Profit Boardroom. There's 3,000 members running the same stacks, so someone has already solved this. I read every question personally and reply with the actual fix, like you've seen today. And the answer becomes a tutorial that's stored forever.

[00:26:25]Coaching calls, um every week, are inside the AI Profit Boardroom, so you can set things up live. And the wins compound, right? Yours and everyone else's.

[00:26:33]So, if you want the full Agent Operating System, if you, you know, if you want to work with us, then then feel free to check that out, right? And inside the AI Profit Boardroom, you get the full Agent Operating System, so Hermes Cloud, Open Cloud, Notebook LM, and Obsidian all wired into one dashboard, plus the community itself. Then you have 3,000 builders inside that running the same stack, helping each other ship faster, right? And you know, the 13 questions that I've answered inside this guide are just from the last 24 hours alone, right? So, every day looks like this.

[00:27:06]Members ask, members answer, I weigh in, and every answer becomes part of the collective brain, right? Also, what you also get is the Agent Operating System uh zip file ready to deploy.

[00:27:16]You have 3,000 fellow builders, so your questions get answered by members who already solved them, not just me. You have weekly coaching calls where the room troubleshoots together in real time. You get daily tutorials built from the best questions and answers in the community. A 30-day Agent Operating System roadmap, the exact one above to get you running. A member maps you can connect with people uh who are using Agent Operating Systems near you. And we've also got 161 pages of testimonials as you can see right here. So, you know, if you've never done this stuff before, or if you're not technical, it's okay, right? We're all in the same boat. We're all learning together, and that's the aim of the game, right? We're all just learning and growing and helping each other, and that's the best thing about all this stuff. So, that's what the AI Profit Board is about.

[00:28:02]Link in the comments description, or go to the AI Profit Board dot com.

[00:28:06]Let's see what we've got on the questions here.

[00:28:18]Hm. Mana says, "How do you add Codex as well in the Agent Agent OS instead of Open Core alongside Antigravity?" So, what you would do is you would go to whatever you've used to build it, right? And you would just say, "Okay, plug in Codex into this system as well, right?" And you would give it the documentation on the CLI, and then ask it to go from there. So, you can plug it inside there. That's how I would approach it.

[00:28:45]What else we got here? Mana says, new update of anti-gravity is much better.

[00:28:49]Yeah, I think the I I really like the new version of anti-gravity. I think people are just they wished that they you know, Google didn't get rid of VLD, but for most people they're not going to use it anyway.

[00:29:01]How to create an agent operating system?

[00:29:04]So, you can just go to the agents and ask it to build. That's all self, but if you want the full setup for me, you can get it inside here with the zip file and everything else, right? And the prompts and everything else. So, if you want to get that full system, it's all inside the AR Profit Bottom Community.

[00:29:28]Thanks for contributing to the world to get on the track. What's your opinion about Google? I think Google have done a fantastic job, you know, like a couple of years ago people are writing them off. Now, Google are clearly like winning the race, you know, like they're top two, top three in the world, so you know, they're they're doing a fantastic job. They're they're doing very well, right? Like it's pretty amazing what they've achieved with all this stuff.

[00:29:55]What's the best open source AI model?

[00:29:58]For Hermes, um probably Qwen. You know, Qwen 3.7 just dropped. I'd probably go with that. I don't really use open source models that much. I'd rather use the APIs, but you could go with that. What are your thoughts on using the desktop app? I much prefer using my agent operating system simply because I can customize it more, right? It makes me more powerful because I can change it however I want.

[00:30:19]Whereas, for example, if you're using um anyone else's, then you can't customize it in the same way, right? So, that's the way that I like to use the agent operating system instead.

[00:30:30]I don't use local models. I don't use VPS, just use APIs.

[00:30:43]And then, East Coast says hello everybody from Long Island. What's everyone's goal?

[00:30:47]Nice. Official says hey.

[00:30:50]Shilo says hi as well. Hello to you, too. Love your videos. Thank you very much.

[00:30:55]Do you have Um yeah, we want loads of workflow ideas for AN 10 inside the AR Profit Bootcamp.

[00:31:02]But, you know, like you can ask your agent to create them. I think like AN 10 is becoming less and less relevant. All right. And uh Manash says thank you.

[00:31:10]Happy to help. I've answered all the questions. All the questions inside the community as well. Thanks so much watching. I appreciate as always. I'll see you in the next one. Cheers.

[00:31:17]Bye-bye.