brainstack

Team Brain — Hands-On Tutorial

End-to-end walkthrough: from zero to shared AI memory.

Time: 15–20 minutes
Prerequisites: terminal, curl, jq, and either Docker (local) or a free Supabase account (hosted)

Choose Your Path

Path	Who	Time	Requirements
A. Admin (owner)	Setting up a new crew	10 min	Supabase account or Docker
B. Joiner (CLI)	Joining an existing crew	5 min	Invite from admin
C. MCP-only	AI-first workflow	5 min	MCP client (Cursor, Claude Code)

Pick one path and follow it through. The walkthrough uses a demo Jira key DEMO-1.

Path A: Admin — Create a New Crew

Step 1: Choose your setup

Option 1: Local Docker (fastest for trying out)

cd /path/to/brainstack

# Start local Supabase
supabase start

# Bootstrap with local mode
bash core/scripts/team-brain-api.sh bootstrap \
  --team "Tutorial Crew" --admin "Alice" \
  --local --jira DEMO-1

Option 2: Hosted Supabase (production-like)

Create a project at supabase.com (free tier works)
Get your Project URL, anon key, and DB password from Settings → API

cd /path/to/brainstack

bash core/scripts/team-brain-api.sh bootstrap \
  --team "Tutorial Crew" --admin "Alice" \
  --url "https://YOUR_REF.supabase.co" \
  --anon "eyJ..." \
  --db-url "postgresql://postgres:YOUR_DB_PASSWORD@db.YOUR_REF.supabase.co:5432/postgres" \
  --jira DEMO-1 \
  --write-env

Step 2: Verify setup

bash core/scripts/team-brain-api.sh whoami

Expected output:

{
  "display_name": "Alice",
  "role": "admin",
  "team_name": "Tutorial Crew",
  "invite_code": "ABC123..."
}

Bootstrap prints a share bundle. Copy it:

=== SHARE BUNDLE (copy to Slack/chat) ===
Team: Tutorial Crew
Invite: ABC123DEF456GH78
URL: https://....supabase.co
Anon: eyJ...
Jira: DEMO-1

Keep this safe — teammates need the invite + URL + anon to join.

Step 4: Create a commit-safe pin

bash core/scripts/team-brain-api.sh pin set --jira DEMO-1 --team-name "Tutorial Crew"
cat .team-brain/project.json

This file is safe to commit (no secrets). Teammates can pull it for the Jira key.

Step 5: Start sync and remember something

bash core/scripts/team-brain-api.sh start DEMO-1
bash core/scripts/team-brain-api.sh sync-status DEMO-1

Now save your first memory:

bash core/scripts/team-brain-api.sh remember DEMO-1 research \
  --source-ref "DEMO-1#setup" \
  "Tutorial Crew is using Team Brain for DEMO-1. Alice is admin."

Expected: "result": "inserted" or "result": "deduped" if you run it again.

Step 6: Check it worked

bash core/scripts/team-brain-api.sh recall DEMO-1

You should see your memory in the list.

Checkpoint A complete. Skip to Verification Checklist.

Path B: Joiner — Join an Existing Crew

Your admin shared: invite code, Supabase URL + anon key, and Jira key.

Step 1: Set up environment

cd /path/to/brainstack

# Option 1: Environment variables
export TEAM_BRAIN_SUPABASE_URL="https://....supabase.co"
export TEAM_BRAIN_SUPABASE_ANON_KEY="eyJ..."

# Option 2: Edit project.public.env
# nano supabase/project.public.env
# → Replace placeholders with URL + anon from admin

Step 2: Onboard

# Replace with your invite and name
bash core/scripts/team-brain-api.sh onboard ABC123DEF456GH78 "Bob" DEMO-1 --role member

Expected output includes:

{
  "display_name": "Bob",
  "role": "member",
  "team_name": "Tutorial Crew"
}

Step 3: Verify

bash core/scripts/team-brain-api.sh whoami
bash core/scripts/team-brain-api.sh status

Step 4: Start sync and recall

bash core/scripts/team-brain-api.sh start DEMO-1
bash core/scripts/team-brain-api.sh recall DEMO-1

You should see memories from your admin (Alice).

Step 5: Remember something

bash core/scripts/team-brain-api.sh remember DEMO-1 research \
  --source-ref "DEMO-1#bob-joined" \
  "Bob joined and confirmed Team Brain is working."

Step 6: Breakdown

bash core/scripts/team-brain-api.sh breakdown DEMO-1

This generates a draft epic breakdown from all shared memories.

Checkpoint B complete. Skip to Verification Checklist.

Path C: MCP-Only — AI-First Workflow

For Cursor, Claude Code, or other MCP-compatible clients.

Step 1: Ensure Team Brain MCP is configured

In Cursor, check your MCP settings (.cursor/mcp.json or global):

{
  "servers": {
    "team-brain": {
      "command": "python",
      "args": ["/path/to/brainstack/mcp/team-brain/server.py"],
      "env": {
        "TEAM_BRAIN_SUPABASE_URL": "https://....supabase.co",
        "TEAM_BRAIN_SUPABASE_ANON_KEY": "eyJ..."
      }
    }
  }
}

Step 2: Onboard via MCP

In chat, say:

Use Team Brain MCP to onboard me with invite ABC123DEF456GH78, name "Carol", on DEMO-1.

Or if already onboarded:

Use Team Brain MCP to attach DEMO-1.

Step 3: Start sync

I'm starting on DEMO-1 — start Team Brain sync.

The AI should:

Call start MCP tool
Summarize existing memories
Enter sync mode

Step 4: Let AI remember

Work naturally. When the AI finds something useful, it should call remember:

Found: the API uses JWT tokens from /auth/login endpoint.

AI response:

I'll save this to Team Brain.
[Calls remember MCP tool with source_ref "DEMO-1#auth-api"]

Step 5: Recall context

What does Team Brain know about DEMO-1?

AI calls recall and summarizes.

Step 6: Breakdown

Generate a breakdown for DEMO-1 from Team Brain memory.

Checkpoint C complete. Continue to Verification Checklist.

Verification Checklist

Run through this checklist to confirm everything works.

Basic connectivity

whoami returns your display name and role
status shows CREDENTIALS=... OK

bash core/scripts/team-brain-api.sh whoami
bash core/scripts/team-brain-api.sh status

Sync mode

start DEMO-1 succeeds
sync-status DEMO-1 shows mode: active
Cache file exists: .team-brain/cache/DEMO-1.json

bash core/scripts/team-brain-api.sh start DEMO-1
bash core/scripts/team-brain-api.sh sync-status DEMO-1
ls -la .team-brain/cache/

Memory operations

remember inserts a new memory
remember with same content returns deduped
remember with same source_ref but new content returns updated
recall returns memories

# Insert
bash core/scripts/team-brain-api.sh remember DEMO-1 note \
  --source-ref "DEMO-1#test1" "Test memory 1"

# Dedupe (same content)
bash core/scripts/team-brain-api.sh remember DEMO-1 note \
  --source-ref "DEMO-1#test1" "Test memory 1"

# Update (same source_ref, new content)
bash core/scripts/team-brain-api.sh remember DEMO-1 note \
  --source-ref "DEMO-1#test1" "Updated test memory 1"

# Recall
bash core/scripts/team-brain-api.sh recall DEMO-1 "test"

Cross-user (if you have a teammate)

Teammate’s remember shows up in your recall
Your remember shows up in teammate’s recall

Correction / history

correct updates a memory
history shows prior versions
restore recovers an old version

bash core/scripts/team-brain-api.sh correct DEMO-1 \
  --source-ref "DEMO-1#test1" \
  --was "Old understanding" \
  "Corrected understanding"

bash core/scripts/team-brain-api.sh history DEMO-1 --source-ref "DEMO-1#test1"

Breakdown

breakdown DEMO-1 generates markdown
Output file exists: .team-brain/initiatives/DEMO-1-breakdown.md

bash core/scripts/team-brain-api.sh breakdown DEMO-1
cat .team-brain/initiatives/DEMO-1-breakdown.md

Pending review (#67)

Near-duplicate remember returns redundant_candidate (not stored)
remember --queue returns pending_submitted
Admin pending list shows the submission
Admin pending approve <pending-id> promotes to live memory

# After a memory exists, try overlapping content
bash core/scripts/team-brain-api.sh remember DEMO-1 research \
  "Same finding as an existing memory body."

# Queue override (member)
bash core/scripts/team-brain-api.sh remember DEMO-1 research \
  --source-ref "DEMO-1#slug" --queue "Improved finding for admin."

# Admin inbox
bash core/scripts/team-brain-api.sh pending list DEMO-1
bash core/scripts/team-brain-api.sh pending approve <pending-id>

Cleanup

stop DEMO-1 stops sync mode

bash core/scripts/team-brain-api.sh stop DEMO-1
bash core/scripts/team-brain-api.sh sync-status DEMO-1
# → mode: stopped

Optional: Watch for Peer Updates

During long sessions, keep sync fresh (same guidance as onboarding + Cursor skill #37):

# Background watch (poll mode) — once per long spike, not every command
bash core/scripts/team-brain-api.sh watch DEMO-1 &

# Or with push (requires Realtime migration)
bash core/scripts/team-brain-api.sh watch DEMO-1 --push &

When teammates remember, your cache updates. Agents should also recall after long research blocks / ~every 8–10 turns — without nagging every turn. If sync slept, wake (or start); watch is not a substitute.

Optional: Roles Demo

Admin: audit crew permissions

bash core/scripts/team-brain-api.sh list-members
# → display_name + role only (never api_key)

Admin: rotate invite

bash core/scripts/team-brain-api.sh rotate-invite
# → New invite code printed

Admin: set role

bash core/scripts/team-brain-api.sh set-role "Bob" --role viewer
# Bob can now only recall, not remember

Viewer: try to remember (should fail)

# As a viewer:
bash core/scripts/team-brain-api.sh remember DEMO-1 note "Test"
# → Error: forbidden: viewer role is read-only

Governance: delete poisoned memory (member/admin)

# Member removes bad context (tombstone — audit preserved):
bash core/scripts/team-brain-api.sh delete DEMO-1 --source-ref "DEMO-1#bad-claim"

# Viewer cannot delete:
# → Error: forbidden: delete requires member role

# Peer cache: Engineer B deletes → Engineer A's sync loop / realtime push
# evicts the tombstoned row from .team-brain/cache/DEMO-1.json automatically.

# Confirm gone:
bash core/scripts/team-brain-api.sh recall DEMO-1 "bad-claim"
# → memory no longer in results

# Restore content at same source_ref (undelete via remember):
bash core/scripts/team-brain-api.sh remember DEMO-1 research \
  --source-ref "DEMO-1#bad-claim" "Corrected finding after delete."
# → undeleted: true in RPC response

Feedback engine: redundant memory + admin approval (#67)

# Member: near-duplicate blocked (not stored)
bash core/scripts/team-brain-api.sh remember DEMO-1 research \
  "API uses bearer tokens from /auth/token endpoint."
# → redundant_candidate: true, matches[] shows DEMO-1#api-auth

# Member: queue override for admin review
bash core/scripts/team-brain-api.sh remember DEMO-1 research \
  --source-ref "DEMO-1#api-auth" --queue \
  "API uses OAuth2 device flow (preferred for CLI)."

# Admin: review inbox
bash core/scripts/team-brain-api.sh pending list DEMO-1
bash core/scripts/team-brain-api.sh pending approve <pending-id> --note "Better finding"
# → live memory updated; recall shows new body

# Admin: reject duplicate
bash core/scripts/team-brain-api.sh pending reject <pending-id> --note "Duplicate of Alice"

Troubleshooting

Problem	Solution
`unauthorized`	Re-run `whoami`. If broken, re-onboard with new display name
`initiative not found`	Run `attach DEMO-1` first, or include key in onboard
`member already exists`	Choose a different display name
Empty `recall`	Nobody has `remember`ed yet — add the first memory
`redundant_candidate` on remember	Similar memory exists — recall matches, reuse `source_ref`, or `remember --queue`
`list_pending_memories unavailable`	Admin: apply `20260908120001_team_brain_pending_review.sql`
`rate limit exceeded`	Wait 1 hour or ask admin to check rate limit settings

Next Steps

Beginner onboarding — detailed setup guide
Demo & Office Hours — one-pager for presenting
Memory architecture — how sync + merge work
MCP server — AI tool integration
Example project — fixture for testing

Quick Reference

# Join
bash core/scripts/team-brain-api.sh onboard <INVITE> "Name" <JIRA-KEY> --role member|viewer

# Work session
bash core/scripts/team-brain-api.sh start <JIRA-KEY>
bash core/scripts/team-brain-api.sh remember <JIRA-KEY> research --source-ref "<KEY>#slug" "Finding"
bash core/scripts/team-brain-api.sh remember <JIRA-KEY> research --queue "Override proposal"  # admin reviews
bash core/scripts/team-brain-api.sh recall <JIRA-KEY> "search term"
bash core/scripts/team-brain-api.sh pending list <JIRA-KEY>        # admin
bash core/scripts/team-brain-api.sh pending approve <pending-id>   # admin
bash core/scripts/team-brain-api.sh breakdown <JIRA-KEY>
bash core/scripts/team-brain-api.sh stop <JIRA-KEY>

# Check state
bash core/scripts/team-brain-api.sh whoami
bash core/scripts/team-brain-api.sh sync-status <JIRA-KEY>
bash core/scripts/team-brain-api.sh metrics <JIRA-KEY>

# Cursor chat
"I'm starting on YOU_JIRA_TICKET_HERE — start Team Brain sync."
"Remember: CLI entrypoint is in pkg/scaffold."
"What does Team Brain know about scaffold?"
"Breakdown YOU_JIRA_TICKET_HERE from Team Brain memory."

Tutorial complete. You’re ready to use Team Brain with your crew.

This site is open source. Improve this page.

brainstack

Team Brain — Hands-On Tutorial

Choose Your Path

Path A: Admin — Create a New Crew

Step 1: Choose your setup

Step 2: Verify setup

Step 3: Save the share bundle

Step 4: Create a commit-safe pin

Step 5: Start sync and remember something

Step 6: Check it worked

Path B: Joiner — Join an Existing Crew

Step 1: Set up environment

Step 2: Onboard

Step 3: Verify

Step 4: Start sync and recall

Step 5: Remember something

Step 6: Breakdown

Path C: MCP-Only — AI-First Workflow

Step 1: Ensure Team Brain MCP is configured

Step 2: Onboard via MCP

Step 3: Start sync

Step 4: Let AI remember

Step 5: Recall context

Step 6: Breakdown

Verification Checklist

Basic connectivity

Sync mode

Memory operations

Cross-user (if you have a teammate)

Correction / history

Breakdown

Pending review (#67)

Cleanup

Optional: Watch for Peer Updates

Optional: Roles Demo

Admin: audit crew permissions

Admin: rotate invite

Admin: set role

Viewer: try to remember (should fail)

Governance: delete poisoned memory (member/admin)

Feedback engine: redundant memory + admin approval (#67)

Troubleshooting

Next Steps

Quick Reference