Ongoing monitoring is Anton's: it re-screens every monitored customer on a regular basis, and once more when a customer is added or resumed. A monitoring evaluation is yours: one screen of a subject you already know, sent through the API and answered in the response. Anton does not repeat it for you.
What ongoing monitoring sends you
A clear re-screen is recorded and sends nothing. A match opens a review in your own queue, and the review fires intelligence.review.opened. Subscribe to that event, and work the review like any other.
Screen a subject yourself
use_case is monitoring, and the event type it maps to is monitoring.evaluate. entity is required, and entity.type is person or business. Beyond that, the evaluation needs searchable entity or instrument data: an entity.name, an entity.vault_token_id, or an instrument. Without one it is a 422 validation_error.
Use reference_id for your own identifier of the subject, such as cust_7731. A name sent in name is used in memory and never stored; see PII and vault tokens to send a token instead.
The response is 201 with the evaluation already completed. Only payout_screening answers 202.
curl -X POST \ https://api.antonpayments.dev/v1/intelligence/evaluations \ -H "Authorization: DPoP $ANTON_ACCESS_TOKEN" \ -H "DPoP: $DPOP_PROOF" \ -H "Idempotency-Key: monitor-cust_7731-2026-11-21" \ -H "Content-Type: application/json" \ -d '{ "use_case": "monitoring", "entity": { "type": "person", "reference_id": "cust_7731", "name": "John Doe", "country": "US" }, "metadata": { "run": "2026-11-21" } }'
Idempotency-Key is required, and a retry with the same key returns the first response. Build the key from the subject and the run, as above. A key made only from the subject would return the first run's answer on every later run.
metadata is yours: strings only, up to 50 keys, returned on the evaluation and never sent to screening. Put a run date or batch id there so you can match evaluations to your own records.
Read the result
The evaluation carries decision, score, tier, reasons and required_actions, the same as every use case. Act on a decision covers what each value asks of you.
The evaluation does not echo entity, so nothing on it tells you which subject it was about. Keep the evaluation id beside your own record, or put your subject reference in metadata, which comes back with it.
Find earlier runs
GET /v1/intelligence/evaluations pages your evaluations, newest first. Filter with event_type=monitoring.evaluate for monitoring runs, and with decision (approve, review or decline, comma-separated) for the outcome, and bound the dates with created_after and created_before, which must be RFC 3339 UTC with milliseconds and a Z, such as 2026-11-01T00:00:00.000Z. An unknown value is a 400, never an empty list.
Events
A completed evaluation can reach you as intelligence.evaluation.completed, and a failed one as intelligence.evaluation.failed. A review opened from an evaluation names it in evaluation_id; Work the review queue covers reading it.
When it goes wrong
entity or instrument, or a value that is not valid.intelligence capability.