Proof of Human for Claude agents.

Trust is a participation seal, not identity and not correctness. Claude has no navigator.credentials — use the hosted chat bridge.

How it wires up

1. Your agent mints at POST https://bridge.trust.sanctifai.com/api/v1/attestations. Never send or print a Trust API key.
2. Tell the human: open approve_url in Chrome and confirm with a device passkey.
3. Poll GET …/api/v1/attestations/{id} (or …/wait) until status=completed.
4. Keep certificate_url. That is the seal — a human showed up and sealed the payload.

Mint from the agent

const res = await fetch("https://bridge.trust.sanctifai.com/api/v1/attestations", {
  method: "POST",
  headers: { "Content-Type": "application/json" },
  body: JSON.stringify({
    task_type: "GEN",
    domain: "GEN",
    task_subtype: "Chat approval",
    requesting_agent: "Claude",
    taskData: { summary: "opaque task payload" },
    resultData: { decision: "approved" },
  }),
});
const { attestation_id, approve_url } = await res.json();
// Human opens approve_url in Chrome. Then poll:
// GET https://bridge.trust.sanctifai.com/api/v1/attestations/${attestation_id}/wait