note.md + Hermes

Why this pairing works

Hermes (by Nous Research) is built around persistent, file-based memory and reads AGENTS.md conventions — the same open-agent lineage as OpenClaw, with an emphasis on self-hosted sovereignty. That worldview is note.md's worldview: no hidden state, files as truth, everything inspectable.

Run Hermes over a note.md vault and its accumulated memory stops being an opaque agent artifact and becomes part of your knowledge base: readable in the outline view, linkable with [[wikilinks]], and — crucially — annotatable. You can literally leave margin notes on your agent's memories.

Setup

1. AGENTS.md at the vault root, as always — conventions from llms-full.txt plus your house rules.
2. Configure Hermes's memory/workspace directory to live inside the vault (e.g. agents/hermes/), or have it write its outputs into your vault folders.
3. Let it work. Read what it wrote in note.md; annotate.
4. Instruct Hermes to consult *.note.md sidecars before revisiting a topic — your corrections become its training wheels.

FAQ

Is Hermes the same as OpenClaw?

No — Hermes is Nous Research's open agent focused on persistent memory and self-hosted operation; OpenClaw is a separate viral open-source personal agent. Both speak markdown and AGENTS.md, so both pair with note.md the same way.

Can multiple agents share one vault?

Yes — that's the design. Plain files plus one AGENTS.md means OpenClaw, Codex, Hermes, and Claude can all work the same vault. Keep it in git so every write is attributable and revertible.