Common commands — your daily brainclaw loop

brainclaw ships 66 MCP tools , but the daily loop runs on a small core. Memorize these.

Start working on something

# MCP (canonical)
bclaw_work intent=execute scope="src/auth/" task="Refactor JWT verification"

# CLI fallback
brainclaw session-start --agent claude-code
brainclaw claim create "JWT verify refactor" --scope src/auth/

bclaw_work opens a session, loads the prompt-sized memory context, and creates a claim — three steps in one call.

Read the workspace

bclaw_context kind=memory path="src/auth/"     # decisions/traps/plans relevant to scope
bclaw_context kind=board                        # active claims, plans, handoffs across the workspace
bclaw_context kind=delta since=sess_abc123      # what changed since prior session

Pick the kind that matches the question: memory for “what do I need to know about this scope”, board for “what’s everyone doing right now”, delta for “what changed since I last looked”.

Capture a decision or trap

# MCP (typed)
bclaw_create entity=decision data={text: "OAuth via auth-gateway", author: "claude-code"}
bclaw_create entity=trap data={text: "Don't store JWT secret in .env.example", severity: "high"}

# Quick capture (auto-classifies)
bclaw_quick_capture text="We decided to swap REST for tRPC on the admin surface"

Hand off to the next agent

# Handoffs are not a bclaw_create entity in 1.10 — use the dedicated CLI
# (or the MCP equivalent), which writes the contract block correctly.
brainclaw handoff "JWT refactor 80% done. Token expiry test still failing." \
  --plan pln_xyz \
  --pre-condition "Run npm test before resuming" \
  --post-condition "All auth/*.test.ts pass"

Close out

bclaw_release_claim id=clm_a planStatus=done
brainclaw session-end --auto-release         # CLI: release any open claims

Done. Tomorrow’s session picks up from here via bclaw_context kind=delta since=<this-session-id>.