Hermes Obsidian: One Shared Memory for All Your AI Agents

Hermes Obsidian is the setup I use to give Hermes Agent — and every other agent I run — one shared memory that lives in an Obsidian vault. Hermes remembers things on its own, but my other agents don’t. Claude Code, Codex and OpenClaw each start fresh. With Obsidian in the middle, they all read from and write to the same notes. In the video below I walk through the daily log system, the vault structure, the one mega-prompt that sets it up, and how I turn those memories into content.

What is Hermes Obsidian?

Obsidian is a free note-taking app that stores everything as plain markdown files in a folder on your computer. That folder is called a vault. Because it’s just files, any AI agent that can read and write files can use it. That’s the whole trick.

Hermes Agent already has its own memory. It keeps a small file called MEMORY.md and a profile file called USER.md. But that memory belongs to Hermes. When I switch to Claude Code to build something, or to Codex to review code, those agents have no idea what Hermes learned yesterday. Each agent works alone.

Hermes Obsidian fixes that. The vault becomes the shared brain. Every agent gets a short pointer that says where the vault is and how to use it. Every agent writes what it did into a daily log. And every night, a summary job tidies the day into notes the next session can use.

If you want the funnel-site version of this guide with extra screenshots, see Hermes Obsidian on the AI Profit Boardroom blog.

What you get from Hermes Obsidian

None of this needs a database or a special server. It’s a folder of notes that both you and your agents can read.

How I structure the vault (PARA)

I use the PARA method to organise the vault. PARA stands for Projects, Areas, Resources and Archive. It’s a simple way to sort notes so both people and agents can find them.

The point is that an agent can be told “check Projects first, then today’s log” and it knows where to look. A messy vault gives you a confused agent.

The one mega-prompt setup

You don’t need to build any of this by hand. In the video I use one copy-and-paste prompt, given to an orchestrator agent, that does the setup for you. It creates the folder structure, writes the pointer each agent needs, and sets up the nightly task.

• Creates the vault folders so everything has a home.
• Gives each agent a pointer — a short note in its own instructions saying where the vault is and what to read first.
• Adds a nightly task so each day’s logs get summarised.
• Adds bonus routines like a morning brief and a health check.

This is the same idea I teach for everything in the Agent OS: tell your agent what you want and let it do the setup. You only step in if something looks wrong.

The official Obsidian skill in Hermes

Hermes ships with a bundled skill called obsidian. According to the Hermes docs, it lets Hermes read, search, create and edit notes in your Obsidian vault. So Hermes already knows how to work with your notes out of the box.

The docs also show how to pair it with a scheduled job. Their example runs a daily paper digest at 8am that uses the arxiv skill and the obsidian skill together:

/cron add "0 8 * * *" "<your prompt>" --skill arxiv --skill obsidian --name "Paper digest"

You can use the same pattern for your own daily jobs. The docs also describe a long-running journal pattern: the same profile, a shared folder and a cron job, with an Obsidian vault as the example. One small gotcha from the docs: workspace folder paths must be absolute, so use the full path to your vault.

Where Hermes’s built-in memory fits

Obsidian doesn’t replace Hermes’s own memory. It sits next to it. Hermes’s MEMORY.md has a limit of about 2,200 characters — roughly 800 tokens. That’s enough for key facts and preferences, not for a whole business.

The docs also note that the memory snapshot is frozen at the start of a session. If Hermes learns something new, start a fresh session with /new so it’s picked up. So I keep the short, always-needed facts in MEMORY.md and USER.md, and everything bigger goes in the vault. For a full rundown of memory options see best Hermes Agent memory.

Voice and automatic capture

Once the memory is shared, voice gets a lot more useful. In the video I show Hermes Apollo voice answering questions using the same notes my other agents write to. It feels like talking to an assistant that was in every meeting.

You can also automate capture. Tools like OMI (a wearable that records conversations) and Screenpipe (which records what’s on your screen) can feed notes into the vault. That’s how you start to build your own long-context engine without typing everything in.

Turn memories into content

This is the part most people miss. The vault isn’t just for agents. It’s a record of what you actually worked on, tested and learned. I have a Create section where an agent pulls ideas from recent logs and turns them into posts, video ideas and emails.

Because the ideas come from real work, they’re specific. You’re not asking an AI to invent a topic. You’re asking it to write up something you already did.

Mistakes to avoid

• Don’t save everything. The biggest mistake is dumping every chat into the vault. Save decisions, results and lessons, not noise.
• Don’t point two agents at the same Hermes profile. The docs warn against it. Share memory through the vault or an external memory provider instead.
• Check saves actually happened. The docs note that small local models can say they saved something without calling the tool. Open the file and look.
• Keep the structure simple. A few clear folders beat a clever system nobody follows.

My verdict on Hermes Obsidian

If you run more than one AI agent, the hardest problem isn’t how smart they are. It’s that they forget, and they don’t share what they know. A plain Obsidian vault, a daily log, a nightly summary and one setup prompt fix that. It costs nothing and every agent can use it. For anyone running Hermes alongside Claude Code, Codex or OpenClaw, Hermes Obsidian is the simplest memory upgrade you can make.

About Julian Goldie

I’m Julian Goldie — founder of Goldie Agency and the AI Profit Boardroom. I test new AI agents almost every day for 432K+ YouTube subscribers and run Hermes, OpenClaw, OpenCode, Claude and Codex side by side in my own Agent OS. Claim your free strategy session or join the AI Profit Boardroom.

FAQ: hermes obsidian

What is Hermes Obsidian?

A setup that uses an Obsidian vault as one shared memory for Hermes Agent and other agents like Claude Code, Codex and OpenClaw, using daily logs and nightly summaries.

Does Hermes Agent support Obsidian?

Yes. Hermes ships with a bundled obsidian skill that can read, search, create and edit notes in your vault.

Do I need to set it up by hand?

No. I use one copy-and-paste prompt given to an orchestrator agent, which creates the folders, the pointers for each agent and the nightly task.

Is Obsidian free?

Obsidian is free to download and stores your notes as plain markdown files on your own computer.

Does Obsidian replace Hermes memory?

No. Hermes keeps short facts in MEMORY.md and USER.md. The vault holds the bigger, shared knowledge that every agent can read.

Related reading

• Best Hermes Agent Memory
• Hermes Memory OS
• Agentic OS v2

Last updated October 2026. This page is a living guide to hermes obsidian — the facts here move fast and it is updated as they do.