INITE Brain

Measured conclusion: Deep-tested. L2 outcome: tools_ok.

Long-term memory for AI agents: a bitemporal knowledge graph with cited answers.

Add to your agent

Copy this into your MCP client config:

{
  "mcpServers": {
    "INITE Brain": {
      "url": "https://brain.inite.ai/mcp",
      "transport": "streamable-http"
    }
  }
}

Measured tools (30)

• about_brain read-only

  What this server is and what it can do, before you connect. Reads nothing and stores nothing.

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• how_to_connect read-only

  The two ways to authenticate, and what each is for. Reads nothing and stores nothing.

  Parameters:

  • client — string — Which client you are configuring, if you want the specific instructions.

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• search_knowledge read-only

  Semantic search over the company knowledge graph. Returns entities with their top facts and external references back to the originating verticals. Apply asOf for historical "what was true on X" queries (valid time: a fact brain learned after X still shows). Embedding model on this tenant: set per workspace.

  Parameters:

  • query (required) — string — Natural-language query
  • asOf — string — Valid-time as-of: facts true at this ISO 8601 moment (a retraction dated before it hides the fact; a fact learned after it still shows — not a knowledge snapshot)
  • limit — integer — Max results (default 10)
  • userId — string — Per-user memory scope: results include tenant-global facts plus this user's personal ones; omit for tenant-global only (fail-closed)
  • predicates — array — Filter to these predicates only
  • minConfidence — number

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• search_multi_hop read-only

  Planner-LLM decomposes the query into ≤ maxHops anchored sub-queries; later hops are anchored to the running entity set so the engine never spends compute on candidates already disqualified. Built for questions that combine evidence across turns / sessions, or that require reasoning over multiple entities ("tenants who complained in April AND upgraded after"). Set synthesize=true to get a grounded answer with citations alongside the per-hop trace. Returns finalEntityIds + supportingFactIds (HotpotQA-style evidence chain) so the caller can audit which facts drove the answer. Embedding model on this tenant: set per workspace.

  Parameters:

  • query (required) — string — Natural-language query
  • asOf — string — Valid-time as-of: facts true at this ISO 8601 moment (a retraction dated before it hides the fact; a fact learned after it still shows — not a knowledge snapshot)
  • limit — integer
  • userId — string — Per-user memory scope: tenant-global plus this user's personal facts; omit for tenant-global only
  • maxHops — integer — Hard cap on planner hops (default 3, capped at 5 — beyond that latency dominates)
  • predicates — array — Filter to these predicates only
  • synthesize — boolean — Run the synthesizer over the final entity set and return a grounded answer with citations
  • synthesisGuardrails — string — Override guardrails when synthesize=true: strict closes to null on partial; lenient returns the answer with the verifier verdict; off skips the verifier

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• graph_retrieve read-only

  Resolves the named entities by canonical name, walks their 1-hop neighbourhood over knowledge_edge, and returns facts across seeds ∪ neighbours, optionally filtered by predicate hints. Works from entities named up front and returns what the graph holds around them ("who runs engineering at Acme" — the answer lives on a neighbour, not on Acme itself). Seeds score 1.0, neighbours 0.7; soft-fails to an empty result.

  Parameters:

  • entityNames (required) — array — Canonical names of the entities to anchor on
  • asOf — string — Valid-time as-of: facts true at this ISO 8601 moment (a retraction dated before it hides the fact; a fact learned after it still shows — not a knowledge snapshot)
  • query — string — Free-text fallback used to resolve seeds when a name does not match exactly
  • predicateHints — array — Prefer facts with these predicates (non-matching neighbour facts are dropped, seed facts kept at lower score)

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• synthesize read-only

  Runs hybrid search then feeds the retrieved facts to a generator LLM that produces a citation-bearing answer (each claim ends with [factId]); a verifier LLM then judges whether every claim is supported. Three guardrail modes: strict (default) returns null on partial / unsupported / verifier outage (fail-closed); lenient returns the answer alongside the verifier verdict; off skips the verifier. Returns a direct natural-language answer rather than raw search results. Embedding model on this tenant: set per workspace.

  Parameters:

  • query (required) — string — Natural-language question
  • asOf — string — Valid-time as-of: facts true at this ISO 8601 moment (not a knowledge snapshot)
  • limit — integer — Top-K facts fed to the generator (default 10)
  • userId — string — Per-user memory scope: tenant-global plus this user's personal facts; omit for tenant-global only
  • predicates — array
  • minConfidence — number
  • synthesisGuardrails — string — Guardrail mode (default = SYNTHESIZE_DEFAULT_GUARDRAILS env)

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• memory_diff read-only

  Returns everything brain learned, unlearned, or replaced between two ISO 8601 cursors [from, to). createdFacts = new active facts; retractedFacts = facts marked retracted in-window with no successor; changedFacts = facts that were superseded by another (carries before+after); newEntities = entities created in-window; forgottenEntities = GDPR-erased tombstones; lapsedExpectations = temporary states (ill, away) whose expected end passed in-window with nothing confirming or ending them — worth asking about. Driving use case: "what changed since the last conversation?" Scope with entityIds and/or predicates to narrow the diff to a feature surface. Window is half-open; consecutive diffs over adjacent windows never double-count.

  Parameters:

  • from (required) — string — Inclusive lower bound (ISO 8601)
  • to (required) — string — Exclusive upper bound (ISO 8601)
  • userId — string — Per-user memory scope: results include tenant-global facts plus this user's personal ones; omit for tenant-global only (fail-closed)
  • entityIds — array — Scope to these entities (short or full ids)
  • predicates — array — Scope to these predicates

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• get_entity_profile read-only

  Full profile of one entity: canonical name, type, externalRefs (cross-vertical ids), and active facts. externalRefs point back to the originating vertical, where fresh state can be rehydrated via @inite/api-kit.

  Parameters:

  • entityId (required) — string — Brain entity id (knowledge_entity:...) or short id
  • asOf — string — Valid-time as-of: facts true at this ISO 8601 moment (not a knowledge snapshot)
  • userId — string — Per-user memory scope: results include tenant-global facts plus this user's personal ones; omit for tenant-global only (fail-closed)

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• get_entity_timeline read-only

  Chronological audit of all facts brain has learned about this entity, including retracted ones. Useful for "what did we know when" investigations.

  Parameters:

  • entityId (required) — string
  • since — string
  • until — string
  • userId — string — Per-user memory scope: results include tenant-global facts plus this user's personal ones; omit for tenant-global only (fail-closed)

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• summarize_entity read-only

  Returns a short one-line briefing about the entity — name, type, the most-confident active facts, external refs — suitable for dropping into an LLM context window. Caches in-process (per companyId / entityId / asOf / styleHint) so a hot entity touched across many turns doesn't reload the profile. styleHint='neutral' | 'sales' | 'support' are template-rendered (no LLM call). styleHint='client_llm' opts into MCP SAMPLING: brain asks the connected client (Claude Desktop / agent runtime) to write the one-liner with its own model — zero brain-side OpenAI cost, perfect for self-hosters who don't want brain holding an LLM key. Falls back to neutral template + sampledBy='local_template' when the client doesn't advertise sampling capability. A briefing in place of the full profile, timeline and competing facts.

  Parameters:

  • entityId (required) — string — Brain entity id (knowledge_entity:...) or short id
  • asOf — string — Summarize what was known at this ISO 8601 moment
  • userId — string — Per-user memory scope: results include tenant-global facts plus this user's personal ones; omit for tenant-global only (fail-closed)
  • styleHint — string — Phrasing register — 'neutral' (default), 'sales', 'support', or 'client_llm' (delegate to client-side LLM via MCP sampling; falls back to neutral template if client doesn't support sampling)

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• get_competing_facts read-only

  Returns facts in COMPETING status — those the conflict resolver couldn't auto-supersede because two same-predicate bitemporal facts overlap in valid-time and are too cosine-close within margin. Grouped by (entityId, predicate); 2-fact groups are pairs the resolver left for adjudication, 3+-fact groups are multi-way disagreements escalated for human review. Answers "is this entity already conflicted on this predicate?" and can drive an in-product reviewer queue. asOf filters to disagreements that were live at that moment.

  Parameters:

  • entityId (required) — string — Brain entity id (knowledge_entity:...) or short id
  • asOf — string — Show what was competing at this ISO 8601 moment
  • userId — string — Per-user memory scope: preflight a user-scoped record_fact against tenant-global priors PLUS this user's personal ones. Omit for tenant-global candidates (matches record_fact's userId)
  • predicate — string — Filter to one predicate (e.g. "status", "address")

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• detect_contradiction read-only

  Dry-run preflight against fn::resolve_fact. Answers "if I were to record this fact right now, what would the resolver decide?" without writing to the database. wouldOutcome ∈ {INSERTED, SUPERSEDED, COMPETING, REJECTED}; reasoning explains which rule fired (semantics class, score gap vs margin, cosine threshold, etc); opposingFacts lists the same-predicate priors the resolver would have weighed against. Useful where the cost of a contested write is high (e.g. agent loops that pay an ingest credit). Fidelity: this preflight uses the SEED source-trust table and authority 0 — the live resolver scores opponents with the learned domain-scoped rates (migration 0045) and declared source authority (migration 0046), so predictions can differ for sources with conflict history or a registered authLevel.

  Parameters:

  • entityRef (required)
  • predicate (required) — string
  • object (required) — string
  • validFrom (required) — string
  • sourceVertical (required) — string — Vertical attributed as source (matches record_fact)
  • userId — string — Per-user memory scope: preflight a user-scoped record_fact against tenant-global priors PLUS this user's personal ones. Omit for tenant-global candidates (matches record_fact's userId)
  • confidence — number
  • validUntil — string

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• get_fact read-only

  Read a single fact as stored — predicate (aspect), object (statement), confidence, validity window, source attribution (vertical / recorder / conversationId), and lifecycle state. Retracted facts still resolve (retracted: true) — "why did I stop remembering this" is part of the trust story; only visibility fences turn into not-found. groundingStatus (grounded | ungrounded) appears on facts stamped by the claim-grounding plane (EVIDENCE_GROUNDING_STAMP); legacy rows carry no key. Same shape as GET /v1/facts/:id. Takes a factId as returned by searches and writes and returns the full trust record behind it.

  Parameters:

  • factId (required) — string — Fact id (knowledge_fact:...) or short id

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• get_fact_provenance read-only

  Return the grounding provenance behind one fact: the verbatim conversation turns it was derived from (episodes, chronological, with char-span quotes when the deriver stamped them). When the server runs with the recursive-closure flag the response additionally carries derivedFacts + closure (the transitive derivedFrom support graph); a plain deployment returns episodes only — the response is a passthrough of GET /v1/facts/:id/provenance, no field filtering. Traversal depth is a server-side policy (flag + caps), not a caller knob. Lets a surprising fact be audited before it is trusted or retracted.

  Parameters:

  • factId (required) — string — Fact id (knowledge_fact:...) or short id

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• find_related_entities read-only

  Get entities connected to the given one via the knowledge graph.

  Parameters:

  • entityId (required) — string
  • asOf — string — Bitemporal edge cutoff — connections as they were believed at this ISO 8601 moment (mirrors GET /v1/entities/:id/connections?asOf=)
  • kind — string — Edge kind filter (e.g. "paid_for", "mentioned_in")
  • userId — string — Per-user memory scope: results include tenant-global relations plus this user's personal ones; omit for tenant-global only (fail-closed)

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• match_procedure read-only

  Cosine-matches procedural memory (curated 'how to' patterns the operator recorded) against a free-text context query. Returns top-K procedures sorted by similarity DESC then priority ASC. Surfaces the behaviour rules that apply to a context — e.g. "user asks about pricing" → "mention they're on platinum tier; they get 20% off". Procedural memory is the third tier alongside facts (semantic) and episodes (timeline).

  Parameters:

  • query (required) — string — Natural-language context
  • limit — integer
  • minSimilarity — number — Floor (default 0.4); rows below are dropped

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• list_procedures read-only

  Paginated listing of procedural memory for admin/review UIs. Sorted by priority ASC, createdAt DESC. Set includeRetired=true to see soft-deleted rows.

  Parameters:

  • limit — integer
  • includeRetired — boolean

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• search_communities read-only

  Cosine-matches the query against topic-community summaries — the coarse retrieval scope. A community is a cluster of related entities (label propagation over the entity graph) with one rolled-up summary. Suited to broad questions ("what do we know about X domain"): it returns an overview of a domain rather than individual facts. Communities are (re)built off-hours by the dreams loop; an empty result means clustering has not run or the graph is too sparse.

  Parameters:

  • query (required) — string — Natural-language topic / domain
  • limit — integer
  • minSimilarity — number — Floor (default 0.3); rows below are dropped

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• list_communities read-only

  Paginated listing of topic communities, largest first. Each row carries a label, a rolled-up summary, and the member count. Gives the knowledge map at a glance.

  Parameters:

  • limit — integer

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• find_entity_communities read-only

  Returns the topic communities a given entity is a member of. Places an entity in its broader context — which clusters / themes it participates in — and leads from one entity to the wider neighbourhood the cluster summarises.

  Parameters:

  • entityId (required) — string — knowledge_entity id

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• why read-only

  Return the design decisions, rationale, invariants and gotchas recorded against a code anchor — the non-derivable engineering "why" a parser cannot recover from source. The anchor is a SCIP-style symbol string ("pkg/namespace/symbol") or a file path ("repo/path/file.ts"). Pass asOf to recall what was known at a past instant (bitemporal). Returns found:0 with empty memory when nothing is recorded.

  Parameters:

  • symbol (required) — string — Code anchor — SCIP-style symbol or file path
  • asOf — string — Bitemporal cursor — recall what was known as of this instant
  • userId — string — Per-user memory scope: tenant-global decisions plus this user's own; omit for tenant-global only

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• recall_decisions read-only

  Find recorded design decisions / rationale / invariants / gotchas by NL topic ("why do we resolve facts through one gateway?"), across ALL code anchors — a semantic search over code-memory, rather than one known anchor. Returns matches with their code anchor + kind, ranked by relevance.

  Parameters:

  • query (required) — string — Natural-language topic to recall the "why" for
  • limit — integer

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• get_source_reputation read-only

  Return everything brain knows ABOUT a source (not its facts): the operator-declared identity (type, authLevel — authority in conflict resolution), the learned per-domain agreement rates (domain = predicate; null domain = the global blended rate; rates need >=8 samples before the resolver trusts them over the neutral 0.5), and the reputation-over-time history. sourceKey is `vertical:recorder` (recorder falls back to `_`) — the same key facts carry in trustSnapshot.sourceKey. Shows how much weight a fact whose trustSnapshot looks surprising deserves.

  Parameters:

  • sourceKey (required) — string — Source key, e.g. "rent:tenant_bot" or "code:_"

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• workspace_status read-only

  The tenant this credential writes to (companyId, its display name, whether it is a personal workspace), what is already in its memory (entities, active facts, facts recorded in the last 7 days), how many Domain Packs are installed, for a user-bound connection who the user is in this memory (`user.name` — the name the memory has learned from them, null until it has), and `nextSteps` — the setup actions that would actually change something, derived from that state rather than a stored checklist (including asking the user how to address them, once, and recording it). Answers "which workspace am I writing to?" and explains why memory answers may look emptier than expected.

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• record_fact

  Insert a fact about an entity. Triggers brain conflict resolution (INSERTED / SUPERSEDED / COMPETING / REJECTED). Most facts come from event ingestion; this records one directly.

  Parameters:

  • entityRef (required)
  • predicate (required) — string
  • object (required) — string
  • validFrom (required) — string
  • sourceVertical (required) — string — Vertical name attributed as source (e.g. "rent")
  • userId — string — Per-user memory scope: the fact (and, for a vertical+id entityRef, the entity) belongs to this end-user only — invisible to other users and to requests that don't assert the same userId
  • evidence — array — Supporting observations behind the claim (≤10 typed pointers) — stored verbatim in the fact source; their presence marks the fact grounded (claim-grounding plane)
  • confidence — number
  • validUntil — string
  • expectedUntil — string — For a temporary state with no stated end (ill, travelling, away): when it should be over. Not an end — the fact stays current; answers say it was expected to be over once this passes
  • conversationId — string — Conversation this fact was observed in — names an observation, so the fact counts as grounded (claim-grounding plane)

  Hints: read-only no · destructive no · idempotent not declared · open-world no

• link_entities

  Insert an edge between two entities. `kind` is the edge type — `identity_of` merges the `from` entity into `to` (cross-vertical identity reconciliation), other typed edges (`paid_for`, `mentioned_in`, `worked_with`, …) are surfaced in related-entity reads and contribute to PPR / SubgraphRAG context. Most edges come from event ingestion; this inserts one directly. identity_of rejects a request that names the same end twice and any merge that would close a cycle; declaring an identity that already holds answers `alreadyIdentical: true` with no edge, so a replay is safe.

  Parameters:

  • from (required)
  • to (required)
  • kind (required) — string — Edge type (identity_of | paid_for | mentioned_in | worked_with | …)
  • sourceVertical (required) — string — Vertical attributed as source (e.g. "rent")
  • weight — number

  Hints: read-only no · destructive no · idempotent not declared · open-world no

• retract_fact destructive

  Mark a fact as no longer believed. Cascades to facts derived from this one. Does not delete; the row remains for audit.

  Parameters:

  • factId (required) — string
  • reason (required) — string
  • retractedBy — object — Who initiated the retraction — mirrors the REST RetractFactDto. Omit for the pre-existing default ({source: 'system'}); pass {source: 'human', userId} when relaying an operator/user decision so the audit trail records the real initiator

  Hints: read-only no · destructive yes · idempotent not declared · open-world no

• record_procedure

  Record a 'how to' pattern that is surfaced when a similar context appears later. trigger = the context phrase the rule should match against (e.g. "user asks about pricing"); action = the behaviour to apply (e.g. "mention they're on platinum tier; they get 20% off"). priority orders ties when multiple procedures match the same context (lower is higher priority; default 100). decayHalfLifeDays is a forward hook for v0.2 relevance decay; v1 ignores it at read time.

  Parameters:

  • trigger (required) — string
  • action (required) — string
  • priority — integer
  • sourceKind — string — Audit tag for who recorded this — default operator
  • decayHalfLifeDays — integer

  Hints: read-only no · destructive no · idempotent not declared · open-world no

• retire_procedure destructive

  Mark a procedural memory row as retired (sets retiredAt). A retired procedure is no longer matched or listed by default. For a rule that no longer applies — distinct from a hard delete because the row stays for audit.

  Parameters:

  • procedureId (required) — string — procedural_memory:<tail> or just the tail

  Hints: read-only no · destructive yes · idempotent not declared · open-world no

• record_decision

  Capture the non-derivable engineering "why" about a code location: a design decision, its rationale, an invariant that must hold, or a gotcha. Anchored to a SCIP-style symbol or file path (NOT line numbers — survives edits). Stored as a bitemporal typed fact: re-recording a `decided` / `invariant` for the same anchor SUPERSEDES the prior one (decision evolution over time), while `because` / `gotcha` accumulate. Attach commit / location for provenance.

  Parameters:

  • symbol (required) — string — Code anchor — SCIP-style symbol or file path
  • kind (required) — string — decided = a design decision (supersedes prior); because = rationale; invariant = a constraint that must hold (supersedes prior); gotcha = a non-obvious trap (accumulates)
  • text (required) — string — The decision / rationale / invariant / gotcha text
  • commit — string — Commit SHA this was decided in (provenance)
  • location — string — file:line provenance, e.g. src/x.ts:42
  • validFrom — string — When the decision took effect — defaults to now
  • confidence — number

  Hints: read-only no · destructive no · idempotent not declared · open-world no

Publisher website Repository

Measured by MCPYP probing. Untested fields are omitted, never estimated. Listing is always free; MCPYP never proxies calls or stores user tokens.