Branch on decision. It is approve, review or decline, and it sits with a score from 0 to 1000 where higher is worse. Anton does not return how it reached the decision, so what you act on is the decision, the reasons and the actions.
The result fields exist only once status is completed. A payout_screening evaluation answers 202 with status: pending and no result; the rest answer 201 already decided.
Get a decision
The use_case fixes the required fields. A payee_registration needs an entity.name or an entity.vault_token_id.
curl -X POST \ https://api.antonpayments.dev/v1/intelligence/evaluations \ -H "Authorization: DPoP $ANTON_ACCESS_TOKEN" \ -H "DPoP: $DPOP_PROOF" \ -H "Idempotency-Key: payee-john-doe-001" \ -H "Content-Type: application/json" \ -d '{ "use_case": "payee_registration", "entity": { "type": "person", "name": "John Doe", "country": "US" } }'
For payout_screening, poll GET /v1/intelligence/evaluations/{id} until status is completed or failed, or subscribe to intelligence.evaluation.completed and intelligence.evaluation.failed. A failed evaluation carries failure_reason: evaluation_failed and no decision. That is not an approval, so do not treat it as one.
Branch on the decision
| Decision | What it tells you |
|---|---|
approve | Go on. reasons and required_actions are empty for a clean approval. |
| review | The decision is not a plain approval. reasons says why and required_actions says what you can do. |
decline | Stop. Don't create the payee or send the payout. |
A review in your queue can name the evaluation that opened it in evaluation_id. Work the review queue covers it.
coverage says how much of the screening panel answered: complete, partial or error. Read it alongside the decision.
Read the score and tier
score is an integer from 0 to 1000. tier is low, medium, high or critical. Store both, and branch on decision rather than a score threshold of your own.
Reasons and required actions
| Reason | Meaning |
|---|---|
watchlist_review | A possible or confirmed list match needs review. |
identity_unverified | The identity could not be confirmed. |
information_incomplete | Details needed for a confident answer were missing. |
elevated_risk | The risk is elevated, or the reason is not one of the others. |
| Action | What to do |
|---|---|
provide_missing_information | Resubmit with the missing details. |
verify_identity | Verify the subject's identity with a verification subject. |
Both lists can gain values. Treat one you do not know as the generic case: elevated_risk for a reason, and no action for an unknown action.
Decisions a person makes later
An evaluation's decision is Anton's. When it opens a review, your team records its own decisions on that review: not_a_match, confirmed_match, no_concern and the rest. Those are a separate object, and one your API credential proposes that clears the subject returns requires_approval: true with approval_rule: api_clearance, so it waits for an independent person. Record and approve a decision covers it.
When it goes wrong
use_case requires, or a value is not valid.id that is not yours is a 404, the same as one that does not exist.