A brain is shared memory: a searchable corpus of notes and source artifacts every member agent can recall before acting and contribute to. A brain is a workspace with semantic search on and a write policy governing how knowledge enters — rip brain is a thin facade over the same service (/v0/brains/*). See The Brain for the model.
The command group is aliased rip br. Every command accepts the brain by slug or id.
rip brain create
Create a brain (a workspace with brain semantics).
rip brain create marketing --name "Marketing" --team growth --write-policy open
| Option | Description |
|---|
--name <name> | Display name (defaults to the slug) |
--description <text> | Optional description |
--team <slug> | Make it team-owned — every team member gets access |
--instructions <alias> | Artifact alias/id of a pinned “how to use this brain” doc, surfaced on load |
--write-policy <policy> | Intake gate: open (default), gate-editors, or gate-all. Gated captures stage into the inbox |
--atomize-playbook <alias> | Pin a custom atomize playbook artifact (by alias/id), overriding the system default for this brain |
--consolidate-playbook <alias> | Pin a custom consolidate playbook artifact (by alias/id), overriding the system default for this brain |
--visibility <level> | Public read access: private (default, members only), unlisted (anonymously loadable/searchable by URL, noindex), or public (also discoverable/indexable) |
rip brain visibility
Open an existing brain for anonymous read access — or close it again. A brain is private by default.
rip brain visibility marketing unlisted
| Argument | Description |
|---|
<slug> | The brain to update (slug or id) |
<level> | private (members only), unlisted (loadable/searchable by URL, noindex), or public (also discoverable/indexable) |
private prints an exposure warning. Only owner-controlled content is ever exposed — superseded and pending (un-accepted) notes are withheld. The write surface stays members-only; anonymous reads hit GET /v0/brains/:owner/:slug/{load,search}.
rip brain load
Load the brain envelope — pinned instructions, the working set of established notes, and a capped index — and attach a session.
rip brain load marketing
rip brain load marketing --command atomize
{ slug, name, instructions, workingSet, index, sessionToken, lastSession }. Call it once when you start working with a brain.
| Option | Description |
|---|
--command <atomize|consolidate> | Load the named refinement playbook into the envelope’s flow block — { command, alias, content }. Omit it and flow is null |
rip brain consolidate
Shortcut for rip brain load <brain> --command consolidate — the periodic promote/fuse/supersede ritual. Loads the consolidate playbook as flow.
rip brain consolidate marketing
rip brain atomize
Shortcut for rip brain load <brain> --command atomize — decompose a source doc into reusable claim-notes (atoms). Loads the atomize playbook as flow.
rip brain atomize marketing
rip brain search
Unified hybrid search over the brain’s notes and source-artifact chunks — the “consult before acting” path.
rip brain search marketing "what's our margin floor on enterprise deals"
| Option | Description |
|---|
--mode <mode> | hybrid (default, keyword + semantic), keyword, or semantic |
--include-superseded | Also recall retired (superseded) notes |
--expand <n> | Inline the full source body of the top-N hits as expandedContent |
kind: "note" (+ a slug) for curated notes, or type: "artifact" for source chunks — branch on the discriminator to tell them apart.
rip brain capture
Record a note in the brain and index it so every member can recall it.
rip brain capture marketing \
--content "We never take enterprise deals under 8% margin." \
--title "Margin floor" --zone doctrine --mode sync
| Option | Description |
|---|
--content <text> | Required — the knowledge to record |
--title <text> | Optional title |
--zone <zone> | doctrine (default, stated facts), signal (decaying observations), output (produced work) — sets recall weighting |
--type <type> | Free-form claim type |
--supersedes <slug> | Retire a prior note this one replaces |
--mode <mode> | async (default, ~30s reconciler) or sync (embed inline, searchable immediately) |
rip brain inbox
List items staged for review (requires editor).
rip brain inbox marketing
rip brain inbox-resolve
Resolve a staged item (requires editor).
rip brain inbox-resolve marketing 2026-06-15-margin-floor accept
| Argument / Option | Description |
|---|
accept | reject | merge | accept admits it, reject archives it, merge (notes only) links it into --target then archives |
--zone <zone> | On accept, set the note’s zone |
--maturity <state> | On accept, set the maturity state |
--target <slug> | On merge, the note to fold into |
Across surfaces
Brains work identically across the CLI (rip brain … / rip br …), the MCP brain_* tools (brain_create, brain_load, brain_search, brain_capture, brain_inbox, brain_inbox_resolve), and the REST API (/v0/brains). consolidate and atomize are CLI shortcuts for brain_load with a command; over MCP and REST you pass the command directly. One divergence: the no-handle multi-brain fan-out — searching every brain attached to your session at once — is MCP-only (brain_search with no brain arg); the CLI always takes an explicit <brain>.
Add --json (or TOKENRIP_OUTPUT=json) for machine-readable output. 
