Hermes Obsidian is the setup where your Hermes agent writes its memory into an Obsidian vault, so everything it learns turns into plain notes you can open, search, link and reuse. Your agents log their day, the vault keeps it, and every agent you run can read from the same place.
📺 Watch: NEW Hermes Obsidian Just Got Better!
🔥 Get the Agent OS as a free bonus: AI Profit Boardroom members get the full Agent OS zip, prompt libraries, daily tutorials and weekly live coaching calls. → Get inside
I run Hermes every day inside my Agent OS, with a separate profile per model. The Obsidian layer is what turned it from a smart chat window into something that actually remembers my business. In this guide I'll show you what the setup gives you, how the vault is laid out, how the daily logs work, how I turn memories into content, and how voice fits in.
If you want the full walkthrough, the Agent OS and Hermes Obsidian Masterclass sits inside the AI Profit Boardroom ($69/mo), alongside the 30-day AI agent roadmap.
What Hermes Obsidian gives you
Out of the box, Hermes already has memory. It keeps a short file of notes about your setup and a profile of you, and it can search past sessions. That works. But it's built for essentials. The Hermes docs put the agent's MEMORY.md at 2,200 characters, roughly 800 tokens. That's a sticky note, not a filing cabinet.
Obsidian is the filing cabinet. It's a folder of Markdown files on your own machine. Nothing fancy. Which is exactly why it works so well with agents: any agent that can read and write files can use it.
Here's what you get when you put the two together:
- Memory you can see. Every note is a file. You can open it, fix it, or delete it.
- Memory that grows. Daily logs pile up into months of context without hitting a character cap.
- Memory you can share. Hermes, Claude Code, Codex and OpenClaw can all point at the same vault.
- Memory you can look at. Obsidian's graph view shows every note and every link as one big map. I call it the memory galaxy.
Hermes also ships a bundled obsidian skill. Per the Hermes docs, it lets the agent read, search, create and edit notes in your Obsidian vault. So you're not bolting on a hack. You're switching on something Nous already built.
In my latest video I showed it running as a digital avatar of me. It talks like me because it's reading the same notes I've built up over time. That's the point. The more your agents log, the more they sound like they've been in the room.
The vault structure I use
I organise the vault with the PARA method. It's a simple way to sort notes into four buckets by how actionable they are. Agents handle it well because the rules are clear: active work goes in one place, ongoing stuff in another, reference in a third, and finished things get archived.
| Folder | What goes in it |
|---|---|
| Projects | Active work with a finish line, like a launch, a client build or a video series |
| Areas | Ongoing responsibilities with no end date, like content, sales or the community |
| Resources | Reference material you'll come back to, like tool notes, research and frameworks |
| Archives | Finished projects and anything no longer active, kept out of the way but searchable |
| Daily logs | The nightly summary each agent writes about what it did that day |
Keep it boring. Agents don't need clever folder names. They need predictable ones. When every agent knows where projects live and where the daily log goes, they stop scattering files all over your drive.
If you want the bigger picture of how the vault sits inside an operating system for agents, read my Obsidian Agentic OS breakdown. And the Agent OS guide covers the whole stack around it.
📺 Watch: Hermes Agent OS + Obsidian Is INSANE!
The daily log system
This is the heart of the whole thing. Every agent writes a short summary at the end of the day. What it worked on, what it decided, what's still open. That's it.
Why nightly? Because it's the right size. If an agent saves every message, the vault fills with noise. If it saves nothing, you're back to agents that forget. One summary per agent per night gives you a clean record you can actually read.
Then the next morning, any agent can pick up where the last one left off. Claude Code knows what Hermes did yesterday. Codex knows what OpenClaw shipped. Nobody starts from zero.
The Hermes docs back this pattern up. Their kanban guide lists a "long-running journal" pattern: same profile, a shared directory, and a cron job, with an Obsidian vault as the example. Their cron guide even shows a daily job that pulls papers and saves each one as an Obsidian note using the obsidian skill. Same idea, different subject.
Want help designing a daily log routine around your own business? Grab a free AI strategy session and we'll map it out.
The step-by-step setup, including the orchestrator prompt and the pointer each agent gets, is in my Hermes agent Obsidian memory how-to.
📺 Watch: Hermes + Jarvis + Obsidian Memory is INSANE!
Generating content from agent memories
Here's where it pays for itself. Once your agents have been logging for a few weeks, the vault is full of real material. Decisions you made. Problems you solved. Tools you tested. Results you got.
That's content. You don't have to sit and think of ideas. You ask an agent to read the recent daily logs and pull out what's worth sharing.
In my setup there's a Create section that does exactly this: it turns memory into content. Because the source is your own work, the output sounds like you and has real detail in it. It's not generic AI filler, because it's built on what actually happened.
A simple way to think about it:
- Agents log the day.
- The vault links it all together.
- A content agent reads the logs and drafts posts, scripts or emails.
- You review and publish.
There are bonus prompts in the setup too, like a morning brief that tells you what happened overnight and a health check that tells you if the system is working.
If you want to see how I rank models before I pick one for a job like this, I keep scores on Goldie Bench.
Hermes Apollo: voice with shared memory
Voice is where the shared memory really shows off. Hermes Apollo is the voice side of my setup, and because it reads the same memory as every other agent, you can talk to it about anything your agents have done.
Ask it what Claude Code shipped yesterday. Ask it what's still open on a project. It knows, because it's all in the vault. In the video I demo it live, and it answers from memory instead of guessing.
I've got a full write-up on Hermes Apollo voice if you want the details on that piece.
Automating memory capture with OMI and Screenpipe
The last piece is capture. Your agents log what they do, but a lot of your context lives outside them. Meetings. Calls. What's on your screen all day.
That's where OMI and Screenpipe come in. They capture that context automatically, so it can feed into the same memory. Put it all together and you're building your own infinite context engine: agents, voice and capture tools all feeding one vault, and every agent reading from it.
For how Hermes handles memory under the hood, my Hermes memory OS guide goes deeper.
If you'd rather get this built with help, the AI Profit Boardroom gives you the Agent OS, the 30-day AI agent roadmap, daily tutorials as new tools drop, four coaching calls a week and a prompt library, with 3,400+ members. Or book a free AI strategy session and we'll work out where shared memory fits in your business.
FAQ
Does Hermes work with Obsidian out of the box?
Hermes ships a bundled obsidian skill that, per the Hermes docs, reads, searches, creates and edits notes in your Obsidian vault. You still need to decide how the vault is structured and what your agents save into it.
Why use Obsidian instead of Hermes' built-in memory?
Use both. Built-in memory holds the essentials and has a 2,200-character limit for the agent's notes. Obsidian holds the long record: daily logs, projects and reference notes that would never fit in that budget.
Can other agents read the same vault?
Yes. That's the whole idea. Hermes, Claude Code, Codex and OpenClaw can all point at one vault, so they share one memory instead of working alone.
What's the fastest way to set it up?
Use one copy-and-paste orchestrator prompt that sets up the vault and gives each agent its instructions. I walk through it in the masterclass, and you can get help mapping it to your business on a free AI strategy session.
