Introduction
What gitmind is, and why your agents' memory belongs in a git repository you own.
AI agents forget everything between sessions. You explain the same project background, the same preferences and the same “don’t touch that, the mobile app still uses it” again and again, in every app.
gitmind is one memory shared by all of them. Agents save what’s worth keeping as short Markdown notes, and every AI app you use reads from the same notes:
- claude.ai on the web, Claude Desktop and Claude mobile, and ChatGPT,
through a connector to
https://gitmind.app/mcp. - Claude Code and Codex, through the same endpoint, or locally through
the
gitmindCLI, which keeps a synced clone on your laptop.
What makes it different
- Your memory is a git repository. Every note is a Markdown file and every change, from any app or machine, is a commit. You can clone it, read it in Obsidian, review its history and leave with it.
- Notes are structured, not transcripts. Agents write short, standalone
facts like
- [decision] use Postgres over MySQL #db, link related notes with[[wiki links]], and keep a history when facts change. See Writing notes. - It runs no LLM of its own. The agent you’re already using decides what to save; gitmind stores, indexes, searches and syncs. No extra API keys, no proxy between you and your AI provider.
- Open source. AGPL-3.0, and self-hostable on Cloudflare.
Next: Quickstart. Curious how this differs from built-in memory, Basic Memory or Mem0? See How gitmind compares.