You send a request with a use_case, an entity, and optional instrument or payout context. Anton screens the inputs, scores the risk, and returns a decision. A payout_screening evaluation runs the full panel and is accepted with 202: poll the evaluation or subscribe to intelligence.evaluation.completed. Every other use case is decided before the response, with 201.
Unknown fields are rejected and enum values are fixed. Sensitive values are used in memory for the evaluation, and Anton persists redacted decision evidence, not raw PII.
Authentication
Every Intelligence route requires OAuth and DPoP credentials with the intelligence scope. See Authentication. A request without credentials returns 401. Evaluations persist under your merchant account, and you read them back with the list and get endpoints. Idempotency-Key is required on create. To try the API, use your sandbox credentials.
The account also needs the intelligence capability. AI-compliance-only accounts hold only that one and have no payout surface. Start with the quickstart.
Use cases
| Use case | When to use it | Event type |
|---|---|---|
payee_registration | Registering or screening a person or business payee | payee.create |
instrument_screening | Adding or screening a bank account, wallet, card, or IBAN | instrument.create |
payout_screening | Screening a payout before it is sent | payout.create |
merchant_onboarding | Screening a business during onboarding | merchant.onboard |
monitoring | Generic/manual screening | monitoring.evaluate |
Required shape
{
"use_case": "payee_registration",
"entity": {
"type": "person"
}
}For a meaningful evaluation, include at least one searchable input: entity.name, entity.vault_token_id, instrument.vault_token_id, or the matching instrument fields.
Validate before creating
Use validation to test request shape without creating an evaluation:
curl -sS -X POST "https://api.antonpayments.dev/v1/intelligence/evaluations/validate" \
-H "Authorization: DPoP $ANTON_ACCESS_TOKEN" \
-H "DPoP: $ANTON_DPOP_PROOF" \
-H "Content-Type: application/json" \
-d '{
"use_case": "payee_registration",
"entity": {
"type": "person",
"name": "John Doe",
"country": "US"
}
}' | jqField rules
| Field class | Rule |
|---|---|
| Countries | Uppercase ISO 3166 alpha-2, such as US or CA |
| Dates | Real calendar dates in YYYY-MM-DD format |
| Currencies and assets | Uppercase supported currency or asset codes, such as USD or BTC |
| Instrument payloads | The nested payload must match instrument.type |
The evaluation
The response is an evaluation: id (eval_...), object, livemode, status, use_case, and the result fields below. See the quickstart for a complete example.
| Field | Meaning |
|---|---|
coverage | How much of the screening panel answered: complete, partial or error. |
decision | approve, review or decline. See Decisions. |
score | Risk score from 0 to 1000. Higher is worse. |
tier | The score's band: low, medium, high or critical. |
reasons | Why the decision is not a plain approval. Empty for a clean approval. |
required_actions | What you can do about it. |