Team Brain
Git-native shared AI memory for teams.
Your teammate debugged that issue yesterday. Your team decided on REST over GraphQL last month. Your codebase has conventions that every AI session ignores.
Team Brain fixes this. Record lessons, decisions, and conventions once — they're automatically loaded into every Claude Code session, for every team member. No servers. No accounts. Just git.
The Problem
AI coding agents are single-player:
- Your teammate spent 2 hours debugging a Stripe webhook — Claude doesn't know
- The team agreed to use async/await everywhere — Claude uses
.then() anyway
- You onboarded a new dev — their Claude starts from absolute zero
- Architecture decisions live in Slack threads nobody can find
CLAUDE.md exists, but it's manually maintained and nobody updates it.
The Solution
Team Brain stores team knowledge in .team-brain/ and auto-generates a BRAIN.md that Claude reads every session. Commit it to git. Everyone on the team gets the same context.
.team-brain/
├── BRAIN.md # Auto-generated, Claude reads this
├── conventions/
│ └── always-use-async-await.md
├── decisions/
│ └── 001-rest-over-graphql.md
├── lessons/
│ └── 2026-03-30-stripe-webhook-retry.md
└── knowledge/
└── api-rate-limits.md
Features
Record Lessons
/team-brain learn Stripe webhooks retry 3 times with exponential backoff
Claude captures the context from your current conversation, creates a structured entry, and regenerates BRAIN.md.
Record Decisions (ADR Format)
/team-brain decide Use REST over GraphQL for public API
Creates an Architecture Decision Record with Context, Decision, and Consequences sections.
Add Conventions
/team-brain convention Always use async/await, never .then() chains
Conventions get highest priority in BRAIN.md — Claude sees them first.
Search the Brain
/team-brain recall stripe webhooks
Keyword + fuzzy search across all entries. Returns the most relevant matches with context snippets.
Onboard New Developers
/team-brain onboard
Generates a comprehensive onboarding guide from all team brain entries — conventions, decisions, lessons, and project knowledge in one document.
Cross-Tool Generation
/team-brain sync # Regenerate BRAIN.md
/team-brain sync cursorrules # Also generate .cursorrules (for Cursor)
/team-brain sync agents # Also generate AGENTS.md (universal standard)
Auto-Loading
Team Brain includes a SessionStart hook that automatically loads context at the beginning of every Claude Code session. If any entries are newer than BRAIN.md, it regenerates automatically.
Installation
Quick Install
git clone https://github.com/Manavarya09/team-brain.git ~/.claude/plugins/team-brain
Add the SessionStart hook to your settings
Add to ~/.claude/settings.json:
{
"hooks": {
"SessionStart": [
{
"hooks": [{
"type": "command",
"command": "bash ~/.claude/plugins/team-brain/hooks/load-brain.sh"
}]
}
]
}
}
Initialize in your project
/team-brain init
This creates .team-brain/ in your repo root. Commit it to git.
Usage
Quick Start
# Initialize team brain in your project
/team-brain init
# Add your first convention
/team-brain convention Use TypeScript strict mode everywhere
# Record a lesson from today's debugging
/team-brain learn React useEffect cleanup runs on unmount AND re-render
# Record an architecture decision
/team-brain decide Use Zod for runtime validation at API boundaries
# Commit and push so teammates get the context
git add .team-brain/ && git commit -m "team-brain: add initial conventions and decisions"
git push
For Teammates
# Pull latest team brain entries
git pull
# Context loads automatically on next Claude Code session
# Or manually sync:
/team-brain sync
Search and Recall
/team-brain recall validation # Search for entries about validation
/team-brain recall # Show recent entries
/team-brain status # Show stats and contributor info
Entry Format
Every entry is a markdown file with YAML frontmatter:
