Create an Intelligence Evaluation

Evaluates payee, instrument, payout, merchant onboarding, or monitoring inputs through Anton Intelligence.

The request contract is strict and uses fixed enums so enterprise clients can generate SDKs safely. Direct PII may be supplied over TLS or recovered from a vault token; Anton uses it in memory for the evaluation and persists only redacted decision evidence.

Per-use-case required fields. Beyond the always-required use_case and entity.type, each use case imposes additional requirements (a 422 validation_error otherwise):

• payee_registration — entity.name OR entity.vault_token_id.
• merchant_onboarding — entity.type must be business, plus entity.name OR entity.vault_token_id.
• instrument_screening — an instrument block.
• payout_screening — a payout block, plus searchable entity or instrument data.
• monitoring — searchable entity or instrument data. These conditional requirements can’t be expressed in the static required arrays below, so they are enumerated here.

Dual surface. With OAuth+DPoP credentials, the evaluation is persisted under YOUR merchant account (retrievable via the scoped reads below) and the Idempotency-Key header is REQUIRED. The credential must carry the intelligence scope and the account the intelligence capability — every Anton account has it, including AI-compliance-only merchants. Vault-token enrichment only accepts tokens owned by the calling merchant. Without credentials, the request runs on the anonymous public TEST surface (where enabled) under a synthetic shared merchant — public evaluations are not tied to your account.

Asynchronous money-movement evaluations. On the authenticated (OAuth+DPoP) surface, a payout_screening evaluation runs the full AI panel and can take up to ~20 seconds, so it is accepted asynchronously: the response is 202 Accepted with {evaluation_id, status: "pending"} and a Location header. Poll GET /v1/intelligence/evaluations/{id} (which returns the pending state until the evaluation settles), or subscribe to the intelligence.evaluation.completed / intelligence.evaluation.failed webhook events for push notification. All other use cases — and every call on the anonymous test surface — respond synchronously with 200 and the full evaluation.