MCP Tools Reference

@brain-protocol/mcp — 38 tools via Model Context Protocol.

Compatible with Claude Desktop, Cursor, Claude Code, Windsurf, and any MCP client.

Quick Setup

npx @brain-protocol/mcp --setup claude-desktop
# or: claude-code, cursor, windsurf, snak

bothLocal SQLite + Cloud

cloudCloud mode only

Tool Groups

Knowledge Management (5)Knowledge Graph (5)OpenMemory (5)Decision Intelligence (3)Development (Cloud) (3)Verification & Proof (6)Sync (1)Cognition (V4.1) (10)

Knowledge Management

query_knowledgeboth

Full-text search with category, author, tag, and brain_id filters.

Parameter	Type	Required	Description
q	string	yes	Search query
category	string	no	Filter by category
author	string	no	Filter by author
tags	string	no	Comma-separated tags
limit	number	no	Max results (default 10)
offset	number	no	Pagination offset
brain_id	string	no	Filter by brain UUID

archive_knowledgeboth

Store a new knowledge entry with confidence scoring and tags.

Parameter	Type	Required	Description
content	string	yes	Knowledge content
content_type	string	no	markdown, code, decision, pattern, documentation, git_commit, json
category	string	no	Category
author	string	no	Author name
tags	string[]	no	Tags array
confidence	number	no	Confidence 0-1
brain_id	string	no	Target brain UUID
is_public	boolean	no	Public visibility

update_knowledgeboth

Partial update of an existing entry (content, tags, confidence).

Parameter	Type	Required	Description
id	string	yes	Entry UUID to update
content	string	no	New content
content_type	string	no	New content type
category	string	no	New category
tags	string[]	no	New tags
confidence	number	no	New confidence

delete_knowledgeboth

Remove an entry by ID (soft-archive with lifecycle_state).

Parameter	Type	Required	Description
id	string	yes	Entry UUID to delete

get_contextboth

Full cognitive pipeline — FoA workspace with query expansion, spreading activation, analogical reasoning, and retrieval proof.

Parameter	Type	Required	Description
query	string	yes	Natural language query
budget	number	no	Token budget for context assembly (default 4000)
brain_id	string	no	Filter by brain UUID

Knowledge Graph

search_with_contextboth

Graph-enriched hybrid search — returns results with connected graph neighbors for richer context.

Parameter	Type	Required	Description
q	string	yes	Search query
limit	number	no	Max results (default 5, max 20)
graph_depth	number	no	Graph traversal depth for context (default 1, max 2)
brain_id	string	no	Filter by brain UUID

create_edgeboth

Create a typed relationship between two knowledge entries.

Parameter	Type	Required	Description
source_id	string	yes	Source entry UUID
target_id	string	yes	Target entry UUID
edge_type	string	yes	derives_from, validates, extends, implements, conflicts_with, supersedes, tested_by, documented_by, optimizes, related_to
confidence	number	no	Edge confidence 0-1

get_graphboth

Traverse the knowledge graph from any entry via recursive CTE.

Parameter	Type	Required	Description
entry_id	string	yes	Starting entry UUID
depth	number	no	Max depth (default 3, max 5)

suggest_connectionsboth

Find semantically similar entries to link — powered by embedding similarity.

Parameter	Type	Required	Description
entry_id	string	yes	Entry UUID to find connections for
limit	number	no	Max suggestions (default 5, max 20)
brain_id	string	no	Scope search to brain UUID

get_provenance_chainboth

Trace the full provenance chain of an entry — all ancestors via derives_from, validates, extends edges.

Parameter	Type	Required	Description
entry_id	string	yes	Starting entry UUID
depth	number	no	Max depth (default 5, max 10)

OpenMemory

add_memoriesboth

Store a memory with automatic client app tagging (Mem0-compatible).

Parameter	Type	Required	Description
messages	array	yes	Array of {role, content} messages
user_id	string	no	User identifier
agent_id	string	no	Agent identifier
app_id	string	no	App identifier
metadata	object	no	Additional metadata

search_memoryboth

Semantic + graph search across all memories, filtered by client_app or user_id (Mem0-compatible).

Parameter	Type	Required	Description
query	string	yes	Search query
user_id	string	no	Filter by user
agent_id	string	no	Filter by agent
app_id	string	no	Filter by app
limit	number	no	Max results (default 10)

list_memoriesboth

List recent memories sorted by creation time, newest first (Mem0-compatible).

Parameter	Type	Required	Description
user_id	string	no	Filter by user
agent_id	string	no	Filter by agent
app_id	string	no	Filter by app
limit	number	no	Max results (default 10)

get_memory_statusboth

Status, capabilities, memory count, and did:starknet identity.

Parameter	Type	Required	Description
wallet_address	string	no	Starknet wallet address for did:starknet identity
chain_id	string	no	Starknet chain ID (default: SN_MAIN)
brain_id	string	no	Scope memory count to brain UUID

delete_all_memoriesboth

Delete all memories for a user/agent (Mem0-compatible). Requires confirm=true.

Parameter	Type	Required	Description
confirm	boolean	yes	Must be true to confirm deletion
user_id	string	no	Delete only this user's memories
agent_id	string	no	Delete only this agent's memories

Decision Intelligence

record_decisionboth

Record a structured decision with context, options considered, and rationale.

Parameter	Type	Required	Description
decision	string	yes	The decision made
context	string	yes	Context/situation
options	string[]	yes	Options considered (min 2)
chosen	string	yes	Option chosen
rationale	string	yes	Why this option
chain_id	string	no	Previous decision UUID (creates derives_from edge)
confidence	number	no	Confidence (default 0.85)
tags	string[]	no	Tags

get_decision_chainboth

Traverse the decision graph from a starting decision.

Parameter	Type	Required	Description
entry_id	string	yes	Starting decision UUID
depth	number	no	Max depth (default 3, max 5)

get_memory_guidanceboth

Get relevant past decisions and high-confidence patterns before acting (Retrieval-Augmented Execution).

Parameter	Type	Required	Description
task_description	string	yes	Task the agent is about to perform
limit	number	no	Max results (default 10)
brain_id	string	no	Filter by brain UUID

Development (Cloud)

suggest_patternscloud

Analyze code and suggest applicable design patterns from the knowledge base.

Parameter	Type	Required	Description
code	string	yes	Code to analyze
file_type	string	no	typescript, cairo, python, etc.

detect_antipatternscloud

Analyze code for anti-patterns and potential issues.

Parameter	Type	Required	Description
code	string	yes	Code to analyze
file_type	string	no	File type

architectural_advicecloud

Get architectural recommendations for code based on project context.

Parameter	Type	Required	Description
code	string	yes	Code to analyze
context	string	no	Project context (e.g., starknet defi)

Verification & Proof

provecloud

Anchor a knowledge entry on Starknet L3 with a Poseidon hash.

Parameter	Type	Required	Description
entry_id	string	yes	Entry UUID to anchor

verifycloud

Verify a knowledge entry's on-chain proof.

Parameter	Type	Required	Description
entry_id	string	yes	Entry UUID to verify

archive_giza_proofboth

Archive a Giza zkML proof result as a verifiable knowledge entry.

Parameter	Type	Required	Description
proof_id	string	yes	Giza proof identifier
model_id	string	yes	Giza model identifier
input_hash	string	yes	Hash of inference input
output_value	any	yes	Inference output value
verified	boolean	yes	Proof verified
timestamp	string	yes	ISO timestamp
proof_tx	string	no	Starknet tx hash
description	string	no	Human-readable description

query_giza_proofsboth

Query archived Giza zkML proofs with filters.

Parameter	Type	Required	Description
model_id	string	no	Filter by model
verified	boolean	no	Filter by verification
limit	number	no	Max results

link_proof_to_entryboth

Link a Giza proof entry to another knowledge entry.

Parameter	Type	Required	Description
proof_entry_id	string	yes	Archived proof UUID
target_entry_id	string	yes	Target entry UUID
edge_type	string	no	validates (default), derives_from, related_to

get_proof_chainboth

Get the full proof chain for an entry — retrieval proofs, anchoring history, and verification status.

Parameter	Type	Required	Description
entry_id	string	yes	Entry UUID to get proof chain for

Sync

resolve_conflictboth

Resolve a sync conflict between local and cloud versions of an entry.

Parameter	Type	Required	Description
conflict_id	string	yes	Conflict ID to resolve
resolution	string	yes	keep_local, keep_cloud, or keep_both

Cognition (V4.1)

run_proactive_cognitioncloud

Trigger foresight inference + dialectical synthesis (budget-capped LLM, admin only).

Parameter	Type	Required	Description
force	boolean	no	Override budget cap (default false)

get_proactive_statusboth

Monitor proactive cognition — last runs, foresight edges, dialectical syntheses, LLM budget.

No parameters

get_dialectical_synthesesboth

List dialectical synthesis entries with promotion rate and survival metrics.

Parameter	Type	Required	Description
limit	number	no	Max results (default 10)
status	string	no	Filter by lifecycle status (review, active, promoted)

list_contradictionsboth

List contradictions between knowledge entries with status and type filters.

Parameter	Type	Required	Description
status	string	no	Filter: pending, resolved, dismissed
type	string	no	Filter: factual, opinion, hypothesis
limit	number	no	Max results (default 20)

get_contradictionboth

Full contradiction detail — both entries, evidence passages, resolution history.

Parameter	Type	Required	Description
id	string	yes	Contradiction UUID

resolve_contradictionboth

Resolve a contradiction with a verdict (CONTRADICTORY, COMPLEMENTARY, REDUNDANT, ORTHOGONAL).

Parameter	Type	Required	Description
id	string	yes	Contradiction UUID
verdict	string	yes	CONTRADICTORY, COMPLEMENTARY, REDUNDANT, or ORTHOGONAL
rationale	string	yes	Explanation for the verdict

get_spreading_activationboth

Run spreading activation from seed entries through the knowledge graph (Collins & Loftus 1975).

Parameter	Type	Required	Description
seed_ids	string[]	yes	Seed entry UUIDs (1-5)
max_hops	number	no	Max propagation hops (default 3, max 5)
min_activation	number	no	Minimum activation threshold (default 0.1)
decay	number	no	Decay factor per hop (default 0.5)

find_analoguesboth

Find structural analogues — entries in different domains with similar graph topology (Gentner 1983).

Parameter	Type	Required	Description
entry_id	string	yes	Source entry UUID
limit	number	no	Max analogues (default 5)
min_similarity	number	no	Minimum structural similarity (default 0.3)

submit_feedbackboth

Record explicit feedback on a knowledge entry — upvote, downvote, citation, or correction.

Parameter	Type	Required	Description
entry_id	string	yes	Entry UUID
signal	string	yes	upvote, downvote, citation, or correction
context	string	no	Additional context for the feedback
correction_content	string	no	Corrected content (when signal=correction)

get_feedback_summaryboth

Feedback counts, net signal, and recent events for an entry.

Parameter	Type	Required	Description
entry_id	string	yes	Entry UUID

Summary

Total tools

Both modes

Cloud only