Product Brain

Agent-readable product context for any project. Drop /product into your repo and your AI agents start warm — with user understanding, decisions, constraints, and roadmap context already loaded.

Quick start

npx product-brain init

Run from the root of any project. No prompts. No global install.

What it does

Creates a /product directory with:

product/
  AGENT.md          ← agents read this first
  one-pager.md      ← fill this in first
  users/            ← who you're building for
  requirements/     ← what needs to be built
  decisions/        ← what was decided and why
  constraints/      ← non-negotiables
  experiments/      ← what was tried and learned
  competitive/      ← market context
  metrics/          ← how success is measured
  roadmap/          ← what's in flight and what's next
  templates/        ← custom template overrides

How it works

Agents (Claude, Cursor, Copilot, Gemini) read /product as part of their project context. The AGENT.md file tells them when to read each folder and when to write back — after a decision, after an experiment, when scope changes.

This gives agents persistent product memory across sessions. They don't ask the same questions twice. They don't propose what's already been rejected. They check constraints before proposing solutions.

Getting started

1. Run npx product-brain init
2. Fill in product/one-pager.md
3. Add your first user segment with product/users/_template.md
4. Start building — your agent reads the rest

Philosophy

• Written for agents first, humans second
• No server, no index, no sync — plain markdown files
• Works in any repo, any stack, any team size
• Agents maintain it as a byproduct of working, not as a separate task

License

MIT