Agent CertAgent Cert
Quickstart

From zero to your first signed, anchored receipt.

This page takes you from signup to a verified Decision Receipt in about five minutes. You'll issue one receipt with curl, look it up in the dashboard, and verify it publicly at /r.

1. Create your account

Sign up at agentcert.net/signup using Microsoft, Google, or email. After signing in you'll land on /welcome where a personal API key is generated and shown once. Copy it now — we never display it again.

Your account is provisioned with 30 free Decision Credits — enough for ten certified receipts (3 credits each). After that, you fund credits with CERT. See CERT & credits.

2. Set your API key as an env var

export AGENT_CERT_API_KEY="ait_..."

The base URL for the API is your Agent Cert backend (visible in your account settings or as NEXT_PUBLIC_API_BASE when running locally). All examples below use $AGENT_CERT_API_KEY as the bearer.

3. Issue your first receipt

A Decision Receipt is the output of a single AI agent action. You submit the execution metadata (what the agent was asked, what it did, which tools it called) and Agent Cert evaluates, signs, and anchors it.

curl -X POST https://api.agentcert.net/v1/receipts \
  -H "X-API-Key: $AGENT_CERT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "agent_id": "agent_invoice_001",
    "platform": "bedrock",
    "acting_model": "claude-sonnet-4",
    "user_intent": "Approve invoice if under $5,000 and vendor is approved.",
    "agent_output": "Invoice approved.",
    "tools_used": ["invoice_lookup", "vendor_lookup"],
    "retrieval_summary": "Invoice $4,200; vendor on approved list.",
    "policy_rules": [
      "Invoice approval must be below $5,000.",
      "Vendor must be approved."
    ]
  }'

acting_model records which model made the decision — captured on the receipt for Article 12 traceability, and distinct from the Agent Cert evaluator that scores it. Optional but recommended on every call. (See the full field list, including parent_receipt_id for provenance.)

The response is your Decision Receipt:

{
  "receipt_id": "rcpt_01HXTRUST...",
  "trust_score": 91,
  "risk_level": "low",
  "policy_compliant": true,
  "prompt_injection_detected": false,
  "evaluator_model": "qwen-32b-instruct",
  "signature": "ed25519:...",
  "execution_hash": "sha256:...",
  "evaluation_hash": "sha256:...",
  "anchor_status": "pending",
  "verify_url": "https://agentcert.net/r/?id=rcpt_01HXTRUST...",
  "credits_remaining": 27
}

4. View it in the dashboard

Visit /receipts in your account. You'll see the new receipt with its trust score, risk level, and anchor status. Anchoring to Solana typically completes within a few seconds on devnet.

5. Verify it publicly

Send the verify URL to anyone — no account needed:

https://agentcert.net/r/?id=rcpt_01HXTRUST...

Or paste the receipt ID at /r. The page recomputes the hashes, checks the Ed25519 signature, reads the on-chain Solana memo, and reports one of three results: fully verified, mismatched, or anchoring-in-progress.

Next steps

  • Define a policy so future receipts from your agent are evaluated against your own rules — not just the ones you pass per-call.
  • Connect via MCP if your agent supports the Model Context Protocol (Copilot Studio, Claude Desktop, etc.) — no manual API calls needed.
  • Browse integrations for platform-specific walkthroughs.
  • Full REST API reference with every endpoint, parameter, and response code.