Your agents.One memory.
Every session you have had with Claude Code, in one place on your computer. Search it, read it, and answer “how did I fix this last time?”
Everything your agent did.Still there.
Claude Code already saves every conversation on your computer. Brain reads those files and turns them into a list you can browse and search.
- Lists every session with its project, branch and date
- Full conversation: your prompts, the answers, the tools used
- Shows which files were edited
- Reads your existing history, nothing to set up
- Fast: it keeps a small index so the list opens instantly
# see your latest sessions $ node bin/brain.mjs list 10 # find an old fix $ node bin/brain.mjs search token refresh # browse it all in the browser $ node bin/brain.mjs serve ✓ http://localhost:4711 (this computer only)
Find it in seconds.Not in an hour.
Type a few words. Brain looks through every prompt and answer you have ever had and shows the session and the exact line.
Your agent can lookback too.
Brain ships a Claude Code skill. Ask “how did I fix this last time?” and Claude searches your history and answers with the session and date.
skill/brain/SKILL.md
Copy it into your Claude Code skills folder.
Share a session
Send a link to the whole conversation instead of a summary. Not built yet.
Teams
One shared history for a team, with private by default. Not built yet.
Claude Code today.More next.
Each agent gets its own small reader that turns its history into one common format, so Brain never depends on a single tool.
Dashed means planned, not shipped.
Stays on your computer.Secrets stay hidden.
Nothing is uploaded
Brain only reads files already on your machine. The viewer is reachable from this computer only.
Keys and tokens hidden
API keys, tokens and private keys are masked before anything is shown.
Your history is not changed
Brain never edits or deletes your Claude Code files.
Look back at your work.In one minute.
Needs Node or Bun. Run it from the Brain folder.