Explore MemoryCrux

Everything you need to understand what MemoryCrux does, how it works, and how to get started with organisational memory for your AI agents.

Getting Started

Sign up for VaultCrux

MemoryCrux is a service layer within VaultCrux. Create your VaultCrux account to get started. Your account gives you access to the MCP server, API tokens, and the memory dashboard.

Free tier availableNo credit card required

Connect your MCP client

Add the MemoryCrux MCP server to Claude Desktop, VS Code, Cursor, or any MCP-compatible client. One config line with your API token and you're live.

# claude_desktop_config.json
"memorycrux": {
"url": "https://mcp.vaultcrux.com/memory"
}

Import your knowledge

Connect your OpenAI and Claude accounts for automatic syncing, or upload conversation exports manually. MemoryCrux deduplicates, indexes, and signs everything with cryptographic provenance receipts.

OpenAIClaudeManual uploadGmail (coming)Slack (coming)

Start querying

Your agents can now query memory, check claims, detect staleness, and verify actions. Every tool call is receipted and auditable. Upgrade to unlock constraints and pre-action verification.

query_memory(
topic: "production database boundaries",
confidence_threshold: 0.8
)

Features: 41 MCP Tools

Agent Passport

Primary

Identity layer | anchors every other agent-context surface

The passport is how an agent proves who it is to the platform: a BLAKE3-hashed record tracking principal identity, sponsor lineage, and a 5-tier reputation score (unverified → basic → established → trusted → elite). Sync operations are gated by tier, so agents must build local track record before sharing knowledge.

Passport is one answer in a broader picture. An agent arriving in a new organisation answers six questions about itself, and each question is served by a distinct surface. The passport provides Identity; the other five layers plug in alongside it to give the agent enough context to act responsibly.

Six layers of agent self-knowledge

The passport links to each of these layers to compose a full operating context.

Layer	Question	Source
Identity	Who am I to this platform?	Passport (exists)
Operator	Who do I work for and how do they work?	OperatorProfile (v2.3, shipped)
Directive	What am I here to do, specifically?	Task context / role / resume intent
Rules	What must I not do?	Constraints + Shield grants (exist)
Playbook	How is this work typically done here?	Agent Skills + Operating Rhythms
Continuity	What happened since last time?	resume_session + domain changelog

Knowledge & Memory

Group A | 13 tools

The core memory layer. Semantic search across every source, versioned snapshots at any point in time, automatic staleness detection, cross-platform contradiction alerts, and claim verification with confidence scores. Available on all tiers including free.

query_memory

Semantic search across all memory records with topic, date, and confidence filters.

list_topics

List all topics in the knowledge base, scoped to personal or team.

get_versioned_snapshot

Point-in-time knowledge snapshot: what was known about a topic at any timestamp.

get_audit_trail

Full change history for a topic with signed provenance receipts.

check_claim

Verify a claim against stored knowledge. Returns confidence score and supporting sources.

get_freshness_report

Staleness flags per topic: which knowledge is fresh, stale, or unverified.

get_contradictions

Cross-platform divergence alerts when sources disagree about the same topic.

get_session_context

Session continuity: a "welcome back" payload with journal, stale pins, and alerts.

reconstruct_knowledge_state

Temporal reconstruction: what did the system know at this exact moment? Returns signed snapshot.

issue_passport

Issue an agent passport with identity, sponsor lineage, and a 5-tier reputation system.

get_passport

Return your passport with current tier and receipt count. Auto-upgrades when thresholds are met.

sync_pull

Pull enriched facts from a remote memory instance. Requires basic passport tier (10+ receipts).

sync_push

Push local facts to the platform. Requires established passport tier (100+ receipts).

Decisions & Causality

Group B | 5 tools

Navigate the full decision graph. See what was known when a decision was made, trace causal chains from action to outcome, walk correction histories, and audit decisions that relied on stale context. Powered by CoreCrux's Decision Plane.

get_decision_context

Full decision record: what was known, confidence, tool called, outcome, and supersession status.

get_causal_chain

Navigate the linked graph of decisions that produced or modified an artefact.

get_correction_chain

Walk the supersession graph: what went wrong and what corrected it.

get_decisions_on_stale_context

Decisions made on evidence that has since been superseded. The "you decided on old data" audit.

record_decision_context

Capture a human decision with full context. The "why" record organisations never create.

Constraints & Verification

Group C | 6 tools

Encode senior judgment as machine-checkable boundaries. Declare constraints in natural language, check proposed actions against organisational boundaries, and run a composite pre-flight gate that combines Shield, Engine, Watch, and Constraints into a single go/no-go decision.

declare_constraint

Create an organisational constraint: encode senior judgment as a machine-checkable boundary.

update_constraint

Modify an existing constraint. Append-only: creates supersession event, never mutates.

get_constraints

List active constraints filtered by scope, status, or type.

check_constraints

Check if a proposed action intersects any active constraints using inverted retrieval semantics.

verify_before_acting

Composite pre-flight gate: Shield + Engine + Watch + Constraints = single go/no-go with signed receipt.

suggest_constraint

Agent-proposed constraint: discover boundaries during work and suggest them for human review.

Monitoring & Health

Group D | 3 tools

Real-time drift detection, domain pressure signals, and watch alerts wired from the VaultCrux Watch and Signals subsystems. Know when the world has changed beneath your agent's feet before it acts on stale assumptions.

get_pressure_status

Current pressure level and active signals for an artefact or across the tenant.

get_active_alerts

Watch alerts since a given timestamp, filtered by severity.

get_signals_feed

Signals feed: mechanical status updates on confidence shifts, anchor drift, and lineage changes.

Agent Effectiveness

Group E | 6 tools

Make agents aware of what they don't know. Coverage gap analysis before acting, token-budgeted context briefings ranked by risk, decision checkpoints for long-running sessions, contextual escalation that preserves reasoning state, and machine-queryable capability discovery. The tools that turn reactive agents into self-aware ones.

assess_coverage

Coverage gap analysis: what domains have zero or stale knowledge before acting.

get_relevant_context

Scoped context briefing: task-shaped payload ranked by risk-if-missed, within a token budget.

escalate_with_context

Contextual escalation: preserve reasoning state when handing off to a human.

checkpoint_decision_state

Decision checkpoint: receipted snapshot of session state for resumption.

get_checkpoints

Retrieve decision checkpoints for a session. Resume where you left off.

get_platform_capabilities

Capability discovery: machine-queryable manifest of all MemoryCrux tools and surfaces.

Elicitation & Delegation

Group H | 8 tools

The intake funnel for tacit knowledge and the pre-flight check for delegation. Structured five-layer elicitation extracts the operating rhythms and judgment patterns that senior experts can no longer articulate on their own. Delegation readiness assessment and agent context provisioning replace rotting soul.md/heartbeat.md/user.md files with structured, receipted, machine-queryable packages that refresh when underlying records change. Shipped in v2.3.

run_elicitation

Structured interview: gap-driven extraction of tacit knowledge across five layers.

answer_elicitation

Submit an answer. Triggers coverage re-assessment, output routing, and next question.

get_operator_profile

Retrieve the human operator's working context: role, expertise, preferences, quality bar.

update_operator_profile

Append a new profile version with provenance receipt. Never mutates.

get_operating_rhythms

Temporal behavioural patterns: what runs daily/weekly/monthly and who can own it.

assess_delegation_readiness

Pre-flight for delegation: five-dimension check across profile, rhythms, constraints, knowledge, decisions.

provision_agent_context

Structured, receipted context package for agent consumption. Replaces soul.md/heartbeat.md/user.md.

decompose_expertise

Decompile "I handle marketing" into specific delegatable steps with elicitation signals.