API Reference.
Receipts on demand.

Gate a request through your tenant's policy. If allowed, mints a court-admissible notarization (hybrid post-quantum signature, evidence-hashed) and writes to the immutable evidence graph. If denied, returns the policy reasoning — no notarization is minted.

{
  "tenant_id":        "<your-tenant-id>",
  "decision_id":      "<client-supplied-uuid>",
  "principal_id":     "<the-human-or-system-actor>",
  "agent_id":         "<the-AI-agent-acting>",
  "action":           "READ_EVIDENCE",
  "resource_id":      "<the-resource-being-acted-on>",
  "request_category": "GRAPH_READ"
}

{
  "decision": {
    "decision":  "ALLOW",
    "reasons":   ["action_in_safe_categories"],
    "policy_version": "<tenant-policy-hash>"
  },
  "trustatom": {
    "id":             "ta_<hex>",
    "evidence_hash":  "<sha-256>",
    "signature_b64":  "<ed25519-signature>",
    "pq_signature_b64": "<ml-dsa-65-signature>",
    "minted_at":      "2026-05-06T13:56:40Z"
  },
  "timing": { "total_ms": 87 }
}

{
  "decision": {
    "decision":  "DENY",
    "reasons":   ["policy_constraint_violated"],
    "policy_version": "<tenant-policy-hash>"
  },
  "trustatom": null
}

Verify the hybrid post-quantum signature and evidence hash of a notarization. Verification is independent of the issuer — auditors verify any notarization with the public key alone, no Trust Gate access required. Court-admissible under ESIGN and eIDAS.

{
  "evidence_hash":     "<sha-256>",
  "signature_b64":     "<ed25519-signature>",
  "pq_signature_b64":  "<ml-dsa-65-signature>",
  "receipt_payload":   { "...": "the original receipt JSON" }
}

{
  "ok":  true,
  "ed25519":   { "valid": true },
  "ml_dsa_65": { "valid": true },
  "evidence_hash": { "match": true },
  "verified_at":   "2026-05-06T14:02:11Z"
}

Generate a human-readable explanation for a decision by ID. Returns reasoning + compliance framework alignment.

{
  "decision_id": "<previously-minted-decision-id>",
  "mode":        "why_allowed"
}

Modes: why_allowed · why_denied · constraints · similar_decisions

{
  "ok": true,
  "explanation": {
    "summary": "<natural-language-summary>",
    "policy_rules": ["<rule-name>"],
    "compliance_alignment": {
      "nist_ai_rmf":  ["<applicable-control>"],
      "soc2":         ["<applicable-control>"],
      "eu_ai_act":    ["Article 50 transparency"]
    }
  }
}

Tenant-scoped aggregates — total decisions, allow/deny split, receipt count, principal + agent counts.

Recent decisions with receipt IDs, descending by timestamp. Query params: limit (default 20, max 100).

Denied decisions with policy reasoning. Useful for compliance review + auditor walk-throughs. Query params: limit (default 20, max 100).

Full lineage for a single decision — decision node + receipt + policy verdict + actor chain. The artifact auditors actually ask for.

Time-range filtered decisions. Defaults to last 24 hours. Query params: since_ms · until_ms · limit (default 100, max 1000).

Per-agent decision statistics: total, allow count, deny count, deny rate. Anomaly surface for security review. Query param: limit (default 20).

Compliance-framework coverage distribution — which frameworks are most exercised by your tenant's decisions. Useful for audit prep.

Python and TypeScript libraries. One-line decorator gates a function call through Trust Gate, mints a receipt, and returns the wrapped result. Available to onboarded customers via private GitHub repo.

from cwn_trust_gate import trust_gate

@trust_gate(policy="data_pii", tenant_id="<your-tenant-id>")
def export_customer_records(query: str) -> list:
    return db.query(query)

# Every call is gated, signed, receipted automatically.

Model Context Protocol server — lets Claude / GPT / any MCP-compatible agent invoke Trust Gate directly. Available on the MCP registry for onboarded customers.

Wraps every Stripe Agent Toolkit operation with Trust Gate so AI agent payments produce a dual receipt — one from Stripe (the money moved), one from Trust Gate (the decision was authorised).