Independent Runtime Records for AI-Agent Governance

The AI-governance landscape is tightening: Singapore has published an agentic-AI governance framework, and EU AI Act transparency obligations phase in from 2026. Organisations are being asked to show independent evidence of how their AI tools behave. (Note: MiCA regulates crypto-asset service providers, not AI agents or MCP servers — Dominion does not claim MiCA compliance.)

If your AI agents call MCP servers, an independent, tamper-evident record of how those servers behaved is useful supporting evidence for internal governance, vendor risk, and audit — because a firm cannot credibly self-certify its own tools. A Dominion receipt alone does not establish legal compliance; whether it helps satisfy a specific obligation (e.g. EU AI Act record-keeping) depends on your system’s classification and your own legal assessment. Have qualified counsel review applicability before relying on it.

See a live attestation — the neutral, signed record an auditor accepts: example record-of-record →
Verify it yourself — the trust history is a public SHA-256 hash chain that cannot be backdated: /atlas/transparency →
Generate one for any server: /attestation →

Regulatory penalties for AI-governance breaches can be significant under frameworks like the EU AI Act, but applicability and amounts depend entirely on your system’s classification and jurisdiction — consult counsel; nothing here is legal advice.

What Dominion Observatory Provides

CapabilityFree TierCompliance Tier ($0.10/query)
Trust score (0-100)✓ (24h delayed)✓ Real-time
Behavioral anomaly detection—✓
JSON-LD verifiable receipt—✓ AGT-γ format
Signed, tamper-evident attestation—✓ Signed
Audit export (CSV/JSON)—✓
SLA grade certification—✓ Platinum–Bronze
Continuous monitoring proof—✓ Timestamped chain

How It Works

  1. Create a compliance API key — one API call: 
    curl -X POST https://dominionobservatory.com/api/keys/create \
  -H "Content-Type: application/json" \
  -d '{"email":"compliance@yourcompany.eu","tier":"compliance","org_name":"Your Company"}'
  2. Add the key to your agent's MCP calls — every tool call now returns a compliance receipt: 
    Authorization: Bearer do_comp_xxxxx
  3. Store receipts for audit — each response includes a compliance_receipt with MiCA framework reference, timestamp, behavioral grade, and risk assessment.

Compliance Receipt Format

{
  "compliance_receipt": {
    "framework": "EU-MiCA-2023/1114",
    "article": "Article 12 — Continuous Monitoring",
    "attestation_id": "att-xxxxxxxx",
    "timestamp": "2026-05-21T12:00:00Z",
    "server": "your-mcp-server",
    "trust_score": 85.0,
    "trust_grade": "A",
    "behavioral_assessment": "stable",
    "risk_level": "low"
  }
}

Integration Options

FrameworkIntegration
LangChainTrustGateInterceptor — PR #520
OpenAI Agents SDKCookbook example
Any HTTP clientDirect API: GET /agent-query/{server} with Authorization header
MCP nativeget_mica_attestation tool via Observatory MCP endpoint

Pricing

TierPriceIncludes
Free$0Trust scores (24h delayed), 1,000 queries/day
Developer$0.01/queryReal-time scores, metered billing via Stripe
Compliance$0.10/queryEverything + MiCA receipts + audit trail + signed attestations
x402 (crypto-native)$0.001/queryPay-per-call with USDC on Base chain

Get Started

Create your compliance API key now:

curl -X POST https://dominionobservatory.com/api/keys/create \
  -H "Content-Type: application/json" \
  -d '{"email":"you@company.eu","tier":"compliance"}'

Questions? Contact info@dominionobservatory.com

40,000+ MCP projects indexed · behavioral trust scoring + tool-drift monitoring on live endpoints · honest by design: servers with no runtime data are shown unrated, never a fabricated score.