Documentation
The whole setup, in text.
The same system as the videos, written down. Five guides take an empty folder to a vault an agent can read, write and remember.
- 01
Set up the vault
What an Obsidian vault needs before a coding agent can read it: a folder, a rules file, a daily note template, and the five folders everything else lives in.
- 02
Keep the config outside the vault
How the agent config, scripts and memory are symlinked into the Obsidian vault, why they live outside it, and the path trap that breaks the journal script.
- 03
Give the agent memory
How to give a coding agent memory that survives sessions: one directory of small fact files, an index loaded every session, and a rule for what never gets stored.
- 04
Turn on the Obsidian CLI
How to enable the Obsidian command line interface, the command syntax, and the commands an agent needs to read, search and write notes in the vault.
- 05
Permission the agent once
How Claude Code permissions work: the settings files, allow and deny rules, pattern syntax, and the small set of commands a vault agent actually needs.
Before you start
What the guides assume.
- macOS or Linux with a normal shell (the paths are the only thing that change).
- Obsidian desktop, opened once so it can register a vault.
- Claude Code installed, logged in, and running from the terminal.
- An empty folder you can turn into a vault. Nothing is deleted along the way.