Claude Brain gives Claude Code a photographic memory — in one file

Developers tired of asking an LLM the same question twice, this one's for you. Claude Brain, a new GitHub project it has been reported that surfaced on Hacker News, aims to give Claude Code persistent memory by storing everything in a single file: .claude/mind.mv2. No database. No cloud. No API keys, the repo claims. Imagine not having to explain last week’s auth bug for the hundredth time — relief, finally.

How it works

Install in 30 seconds, the README promises. After a quick plugin add and enabling “mind,” Claude’s session context, decisions, bugs and solutions are auto-captured and auto-injected at session start. The project allegedly supports a 200K context window and exposes commands like /mind stats, /mind search "authentication" and a CLI (memvid) for power users who want direct access to the .mv2 file. One file. Git-friendly. Scp-able. Simple.

Why it matters

This taps into two current trends: gigantic context windows and local-first tooling. Version-controlling an LLM’s memory feels both practical and slightly uncanny. Want to onboard a teammate? Send a file. Want to roll back a mistaken decision? Git commit. It has been reported that the repo pitches privacy and speed as selling points, though those claims remain worth testing in the wild. Either way, for teams sick of goldfish-with-a-PhD chatbots, Claude Brain promises a little continuity — and that, frankly, feels like progress.

Sources: github.com/memvid, Hacker News