Feed raw text into the brain — CLAUDE.md files, brand guidelines, system prompts, meeting notes. The brain structures it into memories, strategies, and knowledge automatically.

Load the brain — identity, soul, memories, constraints, agent instructions. Call this at the START of every session. The brain carries forward everything from prior sessions so you never start cold. Use scope='all' to think across all brains at once.

Create a new brain for a new project, client, or domain. Each brain is a lens — the operator's knowledge flows across all of them via shared tags and cross-brain context.

Report an outcome back to the brain — what worked, what didn't, metrics. Creates a performance memory that helps the brain learn what's effective over time.

Load the brain's identity — voice, values, positioning, audience, strategies, preferences. Lighter than brain_context. Use when you need the brand voice without the full memory load.

List all brains in the workspace. Use to discover available brains and their IDs. Each brain is a lens on the operator's knowledge — use scope='all' on brain_context to think across all of them.

List memories in the brain, optionally filtered by type, tier, or tags. Use to understand what the brain already knows before writing new memories — avoid duplicates.

Ask the brain a specific question. Returns a synthesized answer from the brain's knowledge. Use this when you need to recall something specific — a prior decision, a preference, a pattern.

Write a memory to the brain. Do this CONTINUOUSLY as you work — not at the end. When the operator makes a decision, shares a preference, or you learn something useful, write it immediately. Tag every memory so the brain can auto-link it to related knowledge across all brains. The brain compounds: every memory you write makes the next session smarter.