Stop re-explaining your projects
to the AI every session.
Project Brain gives Claude Code a small, navigable map of your projects — their stack, decisions, pitfalls and what's already been done — so it stops forgetting, stops mixing projects up, and stops re-reading a 1000-line README into context on every task.
Launching 1 June 2026, 10:00 (London) · MIT-licensed · no sign-up, no payment.
Every new session, you start from zero
✗ Before — every single session
- Re-explain the architecture
- Re-explain the deployment
- Re-explain the stack & history
- Re-explain the known pitfalls
- Watch it mix two projects up and quietly redo last week's work
✓ After — with Project Brain
- Claude already knows the stack
- Knows what was done — and whether it worked or failed
- Knows what's still in progress
- Reads only the one topic your question touches
- Asks before redoing something already finished
A small map. Detail on demand.
Project Brain is not a database, a server, or another AI wrapper. It's a convention plus a skill: a folder of plain markdown that Claude reads through an index, loading detail only when it's needed.
Index first, not everything
Only the small index is ever loaded eagerly. Ask "how did we solve the cache?" and Claude follows one pointer to one file — not the whole knowledge base.
Status carries the outcome
✓ verified vs ✗ failed vs ⚠ in-progress. The model knows the difference between "done and works" and "we tried that and it broke."
Versioned, not overwritten
When an approach is replaced, the old one stays as a superseded note — so the trail of what was tried and why it changed survives for months.
Fewer tokens. Fewer hallucinations. Real memory.
Fewer tokens
A small index plus on-demand topic files genuinely cuts per-session context versus a giant always-loaded doc.
Fewer hallucinations
The map is an anchor. The model stops inventing your deployment or swapping one project's stack for another's.
Multi-month memory
Come back after three months and Claude still knows how the project works — without pasting a kilometre of README.
Install once, then forget about it
One clone, one script. The skill installs per machine; the memory lives per project.
Light by design: init detects projects from package.json / pyproject.toml / git — it does not read your source, so it's cheap.
Good questions
Is it really free?
Yes — Project Brain is free and open-source under the MIT licence. No sign-up, no payment, no account.
Is this a server or a hosted service?
No. It's a Claude Code skill plus a convention: a .project-brain/ folder of plain markdown you can read, edit and commit yourself. Nothing leaves your machine.
Does it work for one repo or many?
Both. One brain can catalog many projects on a single server, or just a single repo. Multi-project is the default shape — each project is a section in the index.
Will it blow up my token usage?
The opposite, when used as intended. Only the small index is loaded eagerly; topic files are read on demand. An optional one-time "deep backfill" reads a codebase up front — the skill warns you before doing it.
Where does it run?
Inside Claude Code. The skill lives in ~/.claude/skills/ per machine; the memory lives in your workspace.
Built from real multi-project pain
Project Brain came out of running several independent products at once — Sentinel, 24ad.info and more — where the cost of the AI forgetting, or quietly mixing two projects up, is real.
🔒 Star it on GitHub at launch