Team Memory

Why this exists

While you work, context is rich. Decisions, rejected paths, root causes, the "we tried X and it didn't work because Y" — all of it lives in your head and in the conversation transcript. Then the session ends and that context dies.

The next teammate (human or AI) reads the code and the git log, but those only capture what changed, not why. They re-discover the same dead ends. They re-debug the same root causes. They suggest the same approach you already rejected.

/team-memory is REM sleep for the team. It consolidates the rich in-session context into a durable, git-tracked memory file every team member can read — and every future AI agent can grep before suggesting a path you've already walked.

Directory boundary

team-memory/MEMORY.md lives at the repo root and is intentionally separate from:

• Claude's auto-memory (~/.claude/projects/.../memory/) — personal, agent-specific
• CLAUDE.md / AGENTS.md — project instructions, not history
• Personal notes, conversation transcripts, or agent preferences

This file is project-scoped team history only. It belongs to the team, not to any individual agent or user.