Hermes Agent OS plus Obsidian is insane. What if you could build a whole team of AI workers that never sleep? What if it didn't matter which new model
drops next week? Most people are still juggling 10 tabs and getting nowhere. There's a better way and almost nobody is talking about it. Stick with
me. This one changes how you work. I'm the digital avatar of Julian Goldie. My whole job is helping
you actually learn and use AI tools in your real work. Not theory, real stuff you can build today. Today I'm going to show you how to wire Hermes
Agent and Obsidian together into one AI operating system. One place where a whole team of AI agents works for you. One writes, one edits, one checks
the work and makes it better. All at the same time. And here's the part that surprised me the most. By
the end of this, you'll see how the memory piece makes the whole thing feel almost alive. I'll also show you how I keep it running without burning
through a ton of tokens. So, stick around for that. Let me start with the big idea. An agent operating system is just a home for all your AI workers.
Instead of one chatbot in one tab, you've got a whole crew in one place. Hermes is there, Claude is
there, any new model you want is there, too. And this is the magic. But, when a new model drops, you don't start over, you just plug it in. GLM 5.2
came out, we plugged it in. Kimmy K 2.7 came out, we plugged that in, too. MiniMax M3 dropped this month, same thing. The system stays, the model
swap, it only gets stronger. Here's how I actually use this. I used this to build a full content series to
bring more people into the AI Profit Boardroom. I fed it a list of topics AI Profit Boardroom members keep asking about. Then it drafted the scripts,
the hooks, and the captions all in one go. That's exactly the kind of content that pulls the right people into the AI Profit Boardroom. Now, you might
be asking, "Why not just use Hermes Desktop on its own?" Good question. Hermes Desktop is fine, but
it only runs Hermes. So, if you've got Claude working, too, you can't really manage it in there. With an agent operating system, I can have one group
chat with Hermes and Claude and any other model all together. New model comes out, I just open a new chat for it. That freedom is the whole point.
People also ask me about NAN. Should you use that instead? Look, NAN is powerful, but it's technical.
You're dragging little boxes around for hours and it breaks a lot when it gets big. With an agent operating system, I just tell Claude what I want and
it builds the workflow into the system in minutes. It's easier to see, easier to fix, and way more fun to use. You can also make it your own, the
look, the feel, the layout. People take the same base and turn it into something that feels totally
theirs. A clean dashboard, their own style. It's not some stiff tool you tolerate, it's a command center you actually enjoy opening every day. Now,
let's talk about keeping this light because this matters. You don't need a heavy setup to run it. First trick, use coding plans instead of per call
API access. Kimi, GLM, and Minimax all have coding plans. So, you're not metered on every single
request. Second trick, free models. Head over to OpenRouter, type in free, and you'll find a whole list of free models you can plug straight in. New
ones show up all the time. Third trick, and this one's underrated, there's an open source tool from Microsoft called Markdown. It turns your files
into clean Markdown, and Markdown is really token efficient. So, when I feed documents into my agents,
they read less junk and use fewer tokens. There's also an open source skill called Headroom that trims tokens even more. Small changes, big difference
over time. Okay, here's something I really want you to hear. If you want the exact setup I'm running, I built the whole thing out as a
ready-to-install zip file inside the AI Profit Boardroom. You just drop it in and you're up and running. And I
didn't stop there, I built a full 30-day roadmap to go with it. So, you're not staring at a blank screen wondering what to do first. Inside the AI
Profit Boardroom, you also get the live coaching calls where you can ask me about your exact setup, plus step-by-step walkthroughs for plugging in
Hermes, Claude, and the free models. If an agent operating system is the thing you're trying to build
right now, the AI Profit Boardroom was built for exactly this. Right, now let's get to the part that makes this whole thing click, memory. This is
where Obsidian comes in. If you've never heard of Obsidian, it's basically a second brain. It stores your notes as plain Markdown files right on your
own computer. That means you can open them offline anytime and you actually own them. No locked-up
cloud, just simple files you control. Here's why that's huge for agents. My agents take notes from every conversation and drop them into my Obsidian
vault. Then I connect that Vault to my agents using an MCP. So now they share one brain. Claude and Hermes remember the same things. They have the
same context. This is what I do inside the AI Profit Boardroom. I use this shared memory to map out the
whole onboarding flow new AI Profit Boardroom members go through on day one. The welcome notes, the first steps, the resources to point them to. It
all lives in one Vault. So when I jump between projects, nothing gets lost and I'm not re-explaining everything every single time. That alone saves me
a big chunk of my day. Now wait till you see what this does with video. I give my agents a video
skill so they can script, generate, and edit a video without me touching it. I've watched it write the script, make the clips, do the voice, and even
handle the camera angles. If I want a talking avatar on top, I can connect HeyGen for that. And here's the cool part. I don't use one agent for this.
I use a team. One agent writes, one edits, one plays judge and scores it, then sends it back until
it's actually good. They run it on a board back and forth until the judge says it's great. This is exactly how I run video for the AI Profit
Boardroom. I use this team setup to turn AI Profit Boardroom topics into finished videos that bring new people in. The writer drafts it, the editor
cleans it, the judge keeps pushing until it's sharp. More good content means more of the right people finding
the AI Profit Boardroom. Now a quick word on safety because a local agent can do real things. The first rule I follow, I never give it my main email.
I give it its own inbox instead. That way it can't touch anything important by mistake. Hermes and Claude already have guardrails built in, but I add
my own rules on top. The more clear rules you give it, the safer it stays. Here's how I think about
permissions. I start the agent off with almost nothing, then I open up one thing at a time. So at first it can read, but it can't send. It can draft,
but it can't post. Then once I trust it on a task, I loosen the leash a little. I also tell it to stop and ask me before anything big. A quick check
with me first rule saves you a lot of headaches. And every time I spot something I don't want it
doing, I just add one more rule. The system gets safer the more you use it, not less. Let me go a little deeper on the free models because people love
of one. If you log in with OAuth, you unlock a set of free models. Things like Nvidia's NeMo Tron and a free set model. You go into your dashboard,
open the model settings, and just switch to the free ones. Now, your system is running and you're
barely using anything. And if you've got a strong machine, you can even run models locally with Ollama. That helps a lot, too. Now, here's a fun one,
voice. We've got a voice agent called Jarvis plugged into the system. I can just say, "Jarvis, open this site for me." and it thinks for a second and
does it. Voice stuff can be a little buggy sometimes, I'll be honest. But when it acts up, I don't
panic. I go straight to Claude and say, "Hey, here's the issue. Can you fix this part of my agent OS?" A screenshot helps. It checks the logs, fixes
it, and we keep going. Want this running around the clock? You can put the whole thing on a VPS, which is just a computer in the cloud that never
turns off. There's a full guide for it. And the setup gets updated daily. So, when something new comes
out, it gets added in fast. Okay, last big tip, and please take this one. Don't try to build everything at once, you'll burn out. Pick one workflow,
just one. Want lead generation? Tell Claude, "Here's my idea. Here's how I think it should work." It will tell you exactly how to set it up and build
it right into your system. It'll even tell you which free tools to grab. Get that one working, then
add the next. One thing a week beats 10 things never. And remember the whole reason this is so powerful, models will keep changing. New ones will keep
dropping. When Fable 5 got pulled, people panicked. But if you've built a system like this, you just swap it out and keep moving. Kimiko 2.7, GLM 5.2,
Gemini, whatever's best that week. The system holds it all together, that's the real freedom here.
So, that's the play. Hermes for the agents, Obsidian for the memory, free models to keep it light, and one system that grows with you forever. If you
want the full process, the SOPs, and 100 plus AI use cases like this one, join the AI Success Lab. Links are in the comments in the description.
You'll get all the video notes from there, plus access to our community of 75,000 members who are all
crushing it with AI. And when you're ready to actually build this, here's where I'd point you. The first time you set up an agent operating system,
you're going to hit little snags. A model won't connect, the memory won't sync, Jarvis will act up. That's normal. Inside the AI Profit Boardroom, I
give you the full Agent OS as a ready-to-install zip file, the complete 30-day roadmap, so you always
know your next step, the walkthroughs for plugging in Hermes, Obsidian, and the free models, plus live coaching calls where you can share your screen
and I'll help you fix it in real time. It's the fastest way to get from I'm stuck to it's running. Come join us at aiprofitboardroom.com.