Persistent memory for AI agents and Claude Code
Say it once. Your agents remember.
Corrections, decisions and ways of working come back in every new session, for your agents only.
In your words, with your name.
Each memory keeps your words, your name, the date and where it applies.
“Use pnpm here, not npm.”
Stopped before it breaks your rules.
In Claude Code, a push, delete or deploy against your rules is stopped.
Connect Claude Code or any MCP agent$ git push --force origin main
See it. Undo it. Erase it.
Each person sees what their agents remember, and can undo or erase any of it.
More cases- “Use pnpm here, not npm.”
- “Never force-push to main.”
- “Book trains for trips under four hours.”