Claude + Hermes + Paperclip is INSANE!

Today, we're going to be answering the latest questions around agent operating systems and how to build them using Hermes and Claude and everything else. Now, if you've never built your own agent operating system, or you're not sure what it is, it's basically one way to orchestrate all of your agents together. So you can see, for example, we have Paperclip built in where we can build a full company of AI agents working together and creating stuff all day 24-7. We also have an AI agent mastermind where our AI agents can talk to each other, and when they come up with good ideas, they add that to our pipeline so that we can just build it all in one single click.

And we can go from idea to implementation ASAP. Now, this is a super powerful system, if you know how to build it the correct way, which is what I'm going to cover today in some of the latest questions from the AI Profit Boardroom, our members inside there, and how to build all this amazing stuff like you can see. I mean, we've even plugged in, for example, Open Design into this system earlier today, and the stuff you can build with this is absolutely mind-blowing. Like, look how powerful this stuff is, and you can do this and customize it as much as you want to whatever you want, whenever you want, using this full system.

If I want to make music, I can do it over here. If I want to do SEO, I can do it over here, right? It's way more powerful and way more fun to build and way more interesting than using Hermes directly inside like a desktop or a terminal or whatever, right? You can customize everything exactly how you want it, which is what we're doing inside this video.

I mean, even if you check out, for example, Grok Build, like we've built some interesting stuff out with that GLM 5.2 as well. You can see some of the cool stuff we've built with this too. So it's just a fantastic way to get the most out of your AI agents. And the great thing is, if you're watching this, you are ahead of 99% of people, and it's just a case of implementing it.

So let's talk through some of the latest questions we've got, and we'll start helping you ASAP inside the AI Profit Boardroom. So first one we got here from Colby. So Colby is new to the AI Profit Boardroom, and he's actually set up the AgentOS system. As you can see, you get my full system inside here, but you could build this from scratch as well if you wanted to.

And so he's basically set up the AgentOS on a VPS with Hostinger and then Hermes Core and OpenCore connected, which is pretty amazing. And he got co-work to help him set it up as well. But now that he's got it running, what should he get started with? So he wants to know how to use the AgentOS system for SEO for real estate clients, right?

How can you do that? What do you build, et cetera? So the way that I would approach this, we actually have this SEO system, which you can see over here. And so that's how I would approach it here, right?

If we take a look, you can basically plug in a keyword, plug in a case study, and then deploy that to whatever websites you want. Now, you could customize that for each of your clients. So you could have a separate tab inside each one of these sections for each client that you have. So for example, if I'm trying to do, let's say I'm trying to do content for the Aero Puffer Boarding, I could have a separate tab here.

And then I could have another tab for this website, right? And then we can have all of our websites plugged into the same system at the same time using this SEO workflow. So all you need to do to customize any part of this, and this is exactly what we did this morning. When we were building in a local AI agent called North Minicode into the system, I just explained exactly what I want into Claude.

And then Claude will build and customize the rest of the agent OS. So if you want to add more SEO workflows for each of your clients, I would ask Claude to improve the agent OS and build a separate tab inside the SEO section for each of your clients. From there, you can plug in a keyword, plug in a case study for each one of your clients and generate SEO content for them. So it's kind of like the SEO workflow we already have built in there, but you've expanded it now to each of your clients and you can easily get Claude to build that in by describing what you want.

And then once it's done, just test, iterate it, make sure it's actually working how you want. You could even have separate prompts and separate skills for each of your clients. But the great thing is if they're all inside the real estate industry, then it's easy to generate a similar prompt for each and just make sure the content is unique for each. So this is a question from Ritz, which is like, when you've got the agent OS system and you're trying to install it or trying to use it inside Codex or GPT 5.5 instead of Claude, does it work the same?

And it's literally the same process. So it doesn't matter which agent you install it with because they're all going to use it the same way. So for example, the instructions that are for Claude run exactly the same with Codex or any other API that you use because they're all coding in the same way and they all understand code using the same fundamentals. So whether you're installing this for Claude or Hermes or Codex, it doesn't matter.

You can still use the same system. This is a great question from Greg. So, you know, when we're using agent operating systems or we're using agents like Hermes in general, how do you install and use a skill from somewhere else safely? You know, because if you look at all the problems that happened with OpenClaude previously, it was down to skills, disguised skills, but actually they had bad intent and then people would install them and it would go off and do crazy stuff.

So how do you avoid that when you're trying to install stuff? Here's a process I'd recommend. I'd read the skill MD file yourself. So that's a markdown file with the details of the skill.

That way you can easily check it and read through it and make sure it's legit. Option number two is you can actually get that skill in that markdown file and give it to Claude and Claude can check for it too. Now, once you've checked it and you understand it, and then also Claude checks it and Claude understands it, you don't even need to install that skill that you're getting from GitHub. You can ask Claude to come up with its own skill that's customized to you and what it knows about you.

So that's the third step really. It's just making sure that you build your own skill and then you don't even touch the original skill. So if you want to be safe, that's a three step process I would recommend. This is pretty cool.

So Amanda actually built out an SEO optimized blog running locally with Quen 330B on a Mac Mini M4 Pro, which is pretty impressive because I have a Mac studio and running Quen, some versions of Quen was super slow. So that's pretty awesome. And basically that's a great example of how you can use this. Now we've actually built out the system inside here for local AI.

And basically what this allows you to do inside the agent operating system, and you can build this too if you're watching for inspiration, is you can plug in a local model like we have North Mini Code that just dropped. And then we can build stuff as you can see right here and we can preview it inside the chat. So this is all switched on, ready to go whenever I need it. We can ask it to build stuff.

When it actually builds stuff, we can preview it and then we can see everything that we've built inside the workspace as well, which is pretty amazing because then we can come back to it, we can see what we've built, we can have a look at everything that we've created. And it's pretty amazing that you can do this with local AI. Bear in mind like you could have this running 24-7, building new things, building useful stuff or running for a to-do list as well. And then this is free, it's local, it's private and it can run 24-7 inside your agent operating system.

So great example from Amanda of how you can do this. And also bear in mind like you can use something like North Mini Code. These local models are getting better all the time and plug that into your agent operating system as well. You don't have to use an API either.

Another question that we got from Larry here is what's the difference between the original Hermes desktop and the one from GitHub by Fata, right? So there's two different versions. There's the official Hermes desktop app that actually dropped later and you can install it here. And then you have the open source version from here, as you can see.

So the main difference is number one is that it's a lot smoother to run the official app from Hermes with your Hermes agent and sync it than it is with the GitHub project. Because the Hermes agent desktop from Noose Research is their official app and so they built it and custom made it and understand the mechanics behind Hermes and then build that around this new desktop app. So if you want a smoother experience, I'd go with Noose Research. The other thing that I noticed when I was using Hermes desktop previously, the original, which is Fata's on GitHub, that one is a lot more complex and has a lot more functionality than the new one from Hermes agent desktop.

The new one from Hermes agent desktop feels like it's been designed for non-technical people who have never used Hermes before and just want something quick and easy with a nice UI. So this is Hermes desktop, the original on GitHub. And then if you want to see the official one from Noose Research that came out later, you can see that, right? So it's Hermes desktop versus Hermes desktop.

And if you look at them side by side, like this one is a lot simpler and a lot more simplified. This one is more complex, but also I've seen that it's more buggy, probably because it is more complex, right? So if you want a smoother experience, just go with the Noose Research one. If you want something amazing where you can have all your agents working together, I would go with AgentOS instead, right?

That's way more customizable and more powerful. By the way, if you want to ask me questions like this, so for example, Larry asked this question, within three hours I've already answered him. If you want to ask me questions like this and I create a video tutorial for you, you can post inside the community of the AI Profitable Boardroom and I personally answer these questions on a daily basis. This is a great post from Alex.

So he's talking about basically what he learned from the situation with Fable 5. So Fable 5 was a new version of Chord that came out, then three days later it got pulled and you can't use it anymore, right? And so that creates a big problem for so many people that I know because they had their full systems built around Fable 5 because it was released and we'd never seen a model taken off the market before, right? So here's the way that he's thinking about it, which is just basically making sure you don't have any one single point of failure for a whole system.

And that's a fantastic learning point and a really good point for us all to learn from. This is why the AgentOS is so powerful because if, for example, one agent gets removed, one model goes down, or one CLI even gets dropped, no problem, you can add a different one in. So, for example, as soon as Fable 5 dropped, we added GLM 5.2 and KimiK217 into that along with local models as well. And that way, you don't need to be reliant on one single model and you can be very adaptable.

And I think the whole direction that you've posted right here makes you anti-fragile. So whatever happens next, it just makes you improve, even if it's a difficult situation like the Fable 5 one. I think the learning here as well is like, you want to own the system so that you don't have to worry about the model. So if you have the AgentOS system set up, well, you can customize that as much as you want and build with it as much as you want.

And it doesn't matter about the models or anything outside of your control. So this is an interesting one from Davide. So Davide was posting about, you know, how, for example, OpenRouter have this fusion model, right? And with this fusion model, you can have five agents working together.

And then with those five agents working together, you tend to get a better answer because one judge will fuse all the answers together, right? Here's an example. And that tends to even outperform Fable 5 on benchmarks, according to the research. Always test it yourself.

But could you set that up? And what we've actually done is we've set up the loop system, which is called Loop Engineering with Hermes. So we've created something similar inside the AgentOS, which is Loop Engineering. And essentially what happens is that you have a builder, you state the goal, you define what done looks like.

And that is one single model. Then you have a separate model, which is a judge inside the AgentOS. You can choose a totally different model or API. And after every round, the judge will check if the builder actually completed the task or not.

And if not, they'll critique it, give feedback, and it will loop back around until the job is done properly. And I think that's a really good way to quality control, check things and also improve the outputs of whatever you're building. Now, if you want to see a full tutorial on that, we've actually got a full system inside the AR Profit Boardroom on Loop Engineering. So you can see that right here.

And this is a full guide on Loop Engineering with video tutorial, full step by step process. Another question that we got from Keith is about AgentOS, Obsidian, Hermes and Paperclip, right? Since AgentOS is using Obsidian Memory Galaxy for AgentOS, Hermes and Paperclip, can just one company in Paperclip have a special Obsidian Vault? So could, for example, if you're wondering what Paperclip is, basically, it's where you can have a team of agents like you can see, they can go off and build stuff, they work together as a team.

And then you can see the builds over here inside the AgentOS. Now, the way that I would approach this is that you have a separate project for that team. And then in the custom instructions for each agent, you can say, OK, log everything you do into this specific Obsidian folder. And also when you're pulling out context for us and building stuff, make sure that you focus only on using that Obsidian folder.

And this way, it's focused on that one folder, that one project folder inside the whole Obsidian setup. And it can use the rest for background information, but it's only going to log and update that one single folder. And it's focused on that. I think that's probably the best way.

And you can do that inside the custom instructions of your AI agents. This is a great question. Is NA10 still relevant if you have an agent operating system, quota code and everything else? Here's the way that I would look at it.

If you're a complete beginner and you've never used AI before, NA10 is fantastic for understanding the fundamentals of AI, how it works and how to build workflows. So if you're a complete beginner, no experience, that is a great way to learn. However, if you want to just skip straight to building agent operating systems because you have experience, because you know what you're doing, because you understand these agents already and you don't feel like you need to use NA10, then you can skip straight to setting up the AgentOS. So it depends on your experience and knowledge already.

Next question from Stan. I'm currently using a Windows PC and it seems like Mac is the way to go. Which Mac would you recommend getting on a budget? So for me, I think I've run the agent operating system on MacBook Air, MacBook Pro.

Those are pretty reasonable. And also one thing I recommend to my team who are using AI a lot is having a Mac mini. So if you have like a permanent setup and maybe you have an office where you go and build stuff regularly, I would recommend a Mac mini is pretty good. But if you need something lightweight, just go with a MacBook Air or MacBook Pro.

Cool thing about the agent operating system as well is it's pretty lightweight. So it doesn't matter what you have, you can still run stuff on it. This is pretty awesome. So Marlon shares their agent operating system, like you can see, and they've fully customized theirs exactly how they want it with an amazing mission control dashboard.

Like you can see, you can add like your own logo and stuff inside it as well. Looks absolutely amazing. Really like the Kanban board built out as well. Looks exciting to use.

Love the way you customize it. So the great thing about this is like, you can see that people are building their own agent operating systems and I've seen so many people do this. I know that if I can do it, if everyone inside the group can do it, we can all build this stuff together. And it's a lot of fun as well.

So Gil was asking about a Mac mini M4 Hermes setup. So Gil says here, I am currently running Hermes on my Mac mini using Quen 3.5 and also Grok, but it seems quite limited. So I want the setup shown in the videos and I understand there is a zip file, which I believe is below. Um, but no, this is a recommended setup from scratch.

Yeah. So the best place to get my setup is here. I've posted the link and then inside this link, you can see the video tutorial, the last update date and the resources right there. So that's how you can get it.

You don't need to just build it with Claude. It could be Codex. It could be, um, for example. Yeah.

So it could be Codex. It could be Claude. It could be Hermes, whatever you want to build with, you can set this up. I think I already answered this one before, but basically I would focus on one thing at a time.

So if you want to set up a to-do list, you can do that inside the chat with Hermes agent, or you could ask it to build out an app for you directly for your to-do list so you can come back to and store it inside the dashboard. And then each one of those things. So for example, emails, lead generation, just focus on one automation a week and just work for it. And you can actually just speak to Hermes and say, Hey, here's my idea.

How do I set this up? And it will guide you through the whole process step-by-step. So Chris asking about like, how do you update the agent OS? So to upgrade it from the daily update, you just use the update.md file inside the zip.

So there's a zip file here. It's got an update.md file in it daily, and you can just use that. By the way, look how fast we answer these questions. Chris asked this 11 minutes ago.

We've already answered him. Pretty wild. All right. Thanks so much for watching.

So if you want to get my full agent operating system, you can get that inside the AI Profit Boardroom, link in the comments description, or go to the AIprofitboardroom.com. This is my AI automation community for helping you save time, grow, and scale with AI automation. Inside the classroom, you can go to the new daily update section and you'll get access to the agent OS system with a video tutorial, the last update date, and also the zip file for installing this. And we also add new daily tutorials based on the new cool stuff that we've added inside there.

So you can learn and grow from every single tutorial we build. If you want to stay updated with it inside the community, you can ask questions, get help and support. And I personally answer them like you've seen today. So you can ask questions inside here and I personally answer every single one of these.

Plus the whole community helps you out too. And there's always people online. You see right now it's 116 people online. So 24 seven, there's people online who you can ask questions to, get help and support from in real time.

And I'll also make video tutorials like this daily answering your questions. So if you want to get all of this, including the classroom with all my best trainings, the agent OS, a four weekly coaching course per week, a map where you can meet people locally who are building with AI agents, just like you. And if you want to be able to DM me, ask me any questions and also post inside the community and just meet lots of cool people and enjoy the journey. You can get it all inside the app offer boarding.

Thanks so much for watching. I'll see you on the next one. Cheers.