Repository navigation
recall: Gemini CLI reads and writes the memory Claude Code keeps for your project #30132
fainir
started this conversation in
Show and tell
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
If you switch between Claude Code and Gemini CLI on the same repo, each one starts the other's sessions knowing nothing. Claude Code keeps notes per project (what to run before saying a change works, where CI lives, how you like answers) as Markdown files under
~/.claude/projects/<project>/memory/. Gemini CLI can't see them.I built recall, an MIT-licensed extension that serves that memory to Gemini CLI over MCP:
It serves the memory of the folder you start Gemini in, with six tools:
memory_list,memory_get,memory_recall,memory_due,memory_writeandmemory_delete. On startup it tells the model to checkmemory_listat the start of a task and save durable lessons withmemory_write, so both agents end up writing to the same notes. The files stay plain Markdown with wiki-links: no database, no vectors, zero dependencies.Since 0.3.0 it redacts secrets (API keys, tokens, passwords in URLs) before a memory leaves the server, unless you start it with
--allow-secrets.Repo: https://github.com/fainir/recall (MIT; also works with Codex, Cursor, Cline and any MCP client).
Honest note: a small side project, not affiliated with Google. Written with help from a coding agent. I'd like to know whether Gemini CLI users keep their own memory files (GEMINI.md or something else) that recall should read too.
All reactions