Agent OS: Local, Obsidian, SEO, NotebookLM MCP + More

So today we're going to be covering all the latest questions on the Agent OS. You can see our mission control right here. And if you've built something like this, or if you're interested in building something like this, you know, for example, we've got a voice activated version of Hermes that we can speak with in real time. We have, for example, the Cloud CLI built in here.

We even have free cloud code and a local engine for building stuff out, plus a powerful memory system. It's basically a way to connect all your AI agents, tools, CLIs in one place, plug them into a memory, and then have custom workflows like you see, for example, we have an agent Kanban over here. We also have, for example, a setup for SEO over here. We can automate videos in one single click, like you can see here, and it's all ready to go.

So if you want a system like this, or you're interested in building something like this, keep watching. I'm going to answer some of the latest questions inside the AI profit boredom community. I'm going to run through these step by step, because I know if people inside the community have these questions, you probably have these questions too. So let's get straight into it.

So question number one from Jose on the prerequisite on AgentOS. For an agent operating system to run locally, what are the hardware requirements and what kind of model should be used? Now, I mean, a lot of people think that you should be using a Mac Mini or a Mac Studio, and I would say that is wrong, because from my own test with a Mac Studio, it's not particularly good unless you've got like the highest spec version for running local models. So I would say, yeah, an RTX is pretty good.

RTX 590 is one of the best setups I've seen. So RTX 590 or a DGX Spark is powerful too. And then for actual models to run, the one that I've seen that seems to be the most powerful is Quen 3.6. That seems to be the most useful to run.

But it also depends what you want to generate. So for example, if you want to generate images locally, well, then you could use something like Ernie. If you just want to generate text, then something like Quen 3.6. So it really depends what you're building with it as well.

The main thing to note here, though, is like if you're just running an agent operating system, it's super lightweight. So you can run this on like a, you know, a non-powerful PC. You could run this on a laptop. If you're not running local models and you're just building an agent OS with APIs or CLIs, then you don't need anything powerful for that.

Skin like Windows setup, WSL or native for agent operating systems. So if you're setting up agent OS on Windows, what should you be using? So I don't run Windows. I have a Mac studio.

But Sheena here answered, and she was saying go native Windows. And she's created some awesome operating systems from what I've seen. So go with the native Windows setup. WSL adds complexity.

And if you're watching this and you figure, okay, which one should I go for for setting up agent OS? Yeah, Windows native, not WSL. So Steve posted here as a welcome, and he's working through the agent OS training as well, taking some notes, organizing some stuff, et cetera. So he was asking about, you know, how to grow his services agency, essentially with an agent OS system.

And there's quite a few use cases for a service-based business that I think you would like with this. So number one is the video agent pipeline. I think you would like that because it will help you market your business better, get more leads, and it's an easy way to automate videos. Number two is you can actually use this for SEO too.

So for example, for us, for our agency website, we use the system inside the agent OS for SEO. So for example, if we scroll down to this section here, you can see that we can plug in a keyword, we can plug in a case study, and then we can deploy that to our websites. And so that's a pretty powerful way to quickly get SEO content ranking. And it actually looks really good when we post it.

You can see an example right here, like nicely formatted, SEO optimized, et cetera. So that is method number two. Also, if you have separate clients and separate businesses that you're working with, you can organize them as separate project folders inside your Obsidian memory. So you can see our Obsidian memory right here.

This is our Obsidian Galaxy. And all of these different stars is a separate memory. So this allows your agents and your CLIs to understand all of your clients, all the projects you're working on. And then if you ask it a question or you ask it to generate some content for a client, it can pull out that information from your Obsidian memory.

There's actually a lot more you can do with it, but those are the top three use cases for using AgentOS. Marlon created their own AgentOS system, as you can see right here. So they've sort of tweaked and customized ours and then built this, which looks pretty awesome. So this is cool because, you know, you can build your own, you can customize your own, you can create it however you want it.

You don't have to have the whole setup like this. So if you don't like the UI or there's something you would tweak, you can easily change this. Here's a good one from Gabrielle. So Gabrielle was talking about how he's ranking inside AI, as you can see right here, for his keyword.

So that's the keyword, and then you can see him ranking right there. It's pretty cool. Awesome job. The AgentOS is great for SEO as well because it can help you rank inside Google, inside ChatGPT, and also inside AR Overviews too.

If you like SEO, you'll probably like that. So, for example, if we type in this keyword here, you can see from our SEO content, we are ranking inside the first sentence here, and then we're also ranking over here, right? And so we're ranking inside Google AR Overviews and also on the first page of Google here, and also here too. And so, you know, I think that it's never been more fun to do SEO and it's never been easier.

And we've built out the systems inside the AgentOS to do that because you can generate videos over here, and you can also generate SEO content over here, and it's pretty powerful. This is an interesting question from Dan, who's asking about how to use NotebookLM and plug it into the AgentOS. So what we actually did is we used an MCP, and that allows us to connect NotebookLM inside our AgentOS. So, for example, we go and check NotebookLM over here.

You can see that we've got all of our notebooks inside this section. We can chat with them. We can do deep research on them. We can speak with them.

We can see our previous conversations with these notebooks as well. And then also we can generate videos or audio views or slide decks or quizzes, data tables, mind maps, infographics, flashcards, and reports directly doing this process. And then once that's done, we can pull them into this section here where we've got everything saved in one place. So it's really easy to organize, and then you can create like amazing research reports or infographics or podcasts or videos using this whole system.

And it's super powerful. And we've got a full system and a guide on it right here, as you can see inside the classroom. So if you go to the classroom, New Daily Updates, you'll find the full tutorial right there. By the way, how fast do we answer these questions?

It's crazy. Like Dan asked this two hours ago. We've already created a video tutorial for him and answered his question directly. So if you want to get your questions answered from me, post inside the AR Profitable Learning community, and I answer these questions every single day inside a video like this.

So Steve was asking about AgentOS, you know, should you use Windows, Linux, Mac? What is everyone running? What operating system are you running on your setup, et cetera? Are you using a VPS?

What do you use? So for me personally, I just prefer to use Mac in general, simply because it's simpler, it's easier, and less things break. And also I noticed a lot of the new releases in AI, they're usually on Mac, which means that if you already have Mac, you're going to get access to the new stuff fastest. Having said that, I think it runs on any setup.

So I've seen people inside the community use a VPS or use Linux or use Windows, and they all get good results. So I think there's no right or wrong way. But for me personally, the simplest and easiest way is Mac. There's a couple of good examples of using VPS and Windows below.

This is a good question. So Khaled is asking about how to use Hermes Workspace from host to VPS. How do you get it set up? Honestly, for me, I've actually found that Hermes Workspace could be quite buggy when you're setting up.

That's why I created the AgentOS instead. For running your setup on a VPS with AgentOS, we've got a full tutorial on it from John, who's already set up his. And he actually talks about how he set it up. And then he also gave his agent operating system access to mobile and desktop as well, which is pretty cool.

So you've got a question from Edwin here, is how does Julian run his AgentOS? Where does he run it? Does he run it on Hostinger or is he using it on a computer that's running 24-7, et cetera? So for me personally, I run it on Mac Studio.

Let me get the settings for this. The reason that I set it up locally and I don't have it hosted is just to keep it safe and sandboxed away from everything else. This is a good question from Abhishek, who's asking about what AI models are we using? Is anyone using, for example, like subscriptions or local models, et cetera?

So I've tested many. So for example, there's some good free APIs out there, like Entu was pretty good. If you want a free API, you can also use News Research Portal with Hermes and then select Step 3.7 Flash or Nemetron 3 Ultra. Owl Alpha is another free API available on OpenRouter.

So there's lots of choices. If you actually go to OpenRouter and type in free, you'll see all the free models they have there too. For actual frontier models, I think the top two right now are Claude Opus 4.8 and also GLM 5.2. I've tested them all and I've actually put them into a benchmark system called GoldieBench, which you can find at goldiebench.com.

So if we check this out, this is basically a website where I've, you know, just based on my own tests. So it's not like an official benchmark or anything like that. But based on my own tests, we grade all the models that we test and then we give them a score. We see how many models have won.

And then you can see, for example, Opus 4.8, GLM 5.2 underneath. Grok underneath that. Grok build was pretty good recently. Quen 3.7, Kimi K2.7.

And then you can compare side by side. So if you're like, I'm not sure, well, you can actually just have a look side by side and see which one you like the most and which one creates the best outputs based on each of these tests, which is pretty cool. Because then you can see, okay, by model, which one is performing the best, which one creates the best stuff, which is pretty amazing. So yeah, I think that's super useful.

And then also you can see the overall scores here. And if you click on one of these pages, you can see the full details of each model, what we built, strengths, weaknesses, what is best for, and then all the demos for everything that we've created for every model. And like, every time I test a new model, this will be updated as a website. This is pretty cool.

So basically what Abhishek has done here is create a system where he's connected Google Search Console, and then he can look at keyword research, and he can run research, and then he can actually use that for analyzing each of his competitors, which is pretty insane. This is actually amazing. I'm inspired to create my own version now. So again, that's just another example of like, if you've got any custom ideas or you want to change or tweak something inside your agent OS, you can see how Abhishek has done it right there.

It's so cool. He's also integrated deployment directly to WordPress and Shopify websites, which is pretty cool as well. This is amazing. So Greg was asking, you know, why would you use N8n instead of agent OS?

Because we cover both inside the AI Profitable Learning. So the reason that we cover N8n for a beginner is like, it's really good for understanding the basics of how automation works and how to set up agents. But if you're more advanced or you have more experience, you can just skip straight to the agent OS system. And you can grab that over here, as you can see.

So here's another question from Greg. Greg was asking about, you know, with the agent OS, should you connect your personal email, et cetera? I personally wouldn't connect your personal email to a system like this. I think it's much better to have a sandboxed email address that your agents get access to and that nobody else knows about.

Because one of the problems that I saw with OpenClaw, particularly when it first came out, is like people connecting their personal emails and then stuff got accidentally deleted. And it's not that secure if you do that. Whereas, for example, if you set up a separate email address and then you give access to only that for the agent, it's still got access to emails. It's still useful, but you don't have all the problems that come with connecting with it to your personal email.

And I think that's much better. So you can see from Jose, he's posted another example of what he's building. Let's take a look at this. Wow.

This is cool. So he's built his own version. I love the memoria. Look at that.

Whoa. This is so cool. Look at the way he's like customized it and built it exactly how he wants. This is what you want, right?

Make it your own. Build what you want to build inside there. That's what I like. Add your own style.

And there's so many people that have done this. So that looks absolutely amazing. He was asking about the notebook LEM setup. So that's the MCP, this one here.

This is interesting. So you can build the agent OS and stuff like this as well, which is like a, yeah, it's like a virtual machine or a VPS and then run your agent OS through that. But also, if you get a good setup with good memory, you could also run local models on that. If you actually type in local inside the community, you can see all the posts on the best local models, how to use them, et cetera.

I think there's a lot of good stuff there. Let's see what we're, so Rhys is asking about paperclip and codex agents. Should codex desktop plugin skills be inherited or use paperclip skills? This is a good question.

I'm actually going to ask the agent that built this to help you. Let's see what else we've got here in the meantime. So Amanda is building this out. Look at that.

This is crazy. I'm so inspired to see all this crazy stuff people are building. Thanks for sharing, Amanda. So we've got another question here from Newman, which is, you know, I travel a lot, always carry my MacBook Air M5 with me.

Since it's lightweight, plan to use it. Should you use a VPS or should you use just a MacBook Air? I would honestly say like the MacBook Air is totally fine for agent OS. It can run the system.

The agent OS itself is pretty lightweight. So unless you're running local models, which 99% of people aren't, you're going to be fine. When I say 99% people, I mean, like most people are not running them effectively, not a frontier level. This is a good question.

So Ritz is asking like, what's the use case of free cloud code compared with Hermes, Agent Factory, cloud, et cetera. So, so the reason that we set up free cloud code inside there is like a lot of people want to use something like cloud, but they don't have the resources or the subscription to get the most out of it. So we set up free cloud code inside there because it's in massive demand and it's quite useful. So if you already have cloud, if you already have Hermes and you're good with that, you don't need to use free cloud code.

Personally, I don't use free cloud code that much, but for some people who want a free version of it, that's available. You can always hide and delete parts from the setup as well, if you ask your agent who set it up to fix that. And that is basically every question answered today. So we've been through the whole community.

I do this every day, answer every single question on a video tutorial to help you as much as I can. I know that if people watching this sort of stuff like agent operating systems, we're probably all got the same questions. If you want to get my setup, you can get it inside the AI Profit Boardroom. Just go to the classroom and then go to Agent OS.

You have a cool thing as well, like you can ask questions and then I give you support and technical help like you've seen today. And I can also ask my agents that set up my Agent OS any questions and then give you the feedback from them as well. So you get the zip file, you get the full guide, the video tutorial, we update it daily too. You also get new trainings when we add new updates inside the system to help you as much as we can.

We have four weekly coaching calls where you meet the community, share your screen, ask the coach any questions. Inside the community, you can post questions and I'll help you daily. And also so does the community. You can see there's 91 people online right now.

There's always people online ready to help you as well. Inside the map, you can meet people locally who are building with AI agents just like you. And this is all inside the AI Profit Boardroom. Link in the comments description or just go to the AIProfitBoardroom.com to get access.

Thanks for watching.