Reviews are your team's own compliance reviews: screenings of yours that need a person, ones opened by monitoring or list screening, and ones your team opens by hand. The portal is where your team works them. The Reviews API lets your own case system read the same queue, and webhooks keep it in step.
The Reviews API and its webhooks need the Reviews API and Webhooks add-on. Without it an endpoint answers 403 module_not_enabled. Every endpoint also needs a credential with the intelligence scope. Anton's own screening of your payouts never appears in a review.
List the queue
GET /v1/intelligence/reviews returns your reviews, newest first. Filter with status, which is open, waiting_on_customer, awaiting_approval or closed, and with kind. Page with limit and cursor; next_cursor is null on the last page. An unknown status, kind or parameter is a 400, never an empty list.
Each row is a review with its number, kind, priority and due_at.
curl https://api.antonpayments.dev/v1/intelligence/reviews?status=open&limit=25 \ -H "Authorization: DPoP $ANTON_ACCESS_TOKEN" \ -H "DPoP: $DPOP_PROOF"
kind is one of possible_sanctions_match, possible_pep_match, screening_review, manual, payout_held_by_rule or payee_risk_by_rule. The last two are opened by your own payout rules, and their money decision is made on the linked approval in the portal, not through this API.
source says where it came from: evaluation, manual, monitoring, list_screening or payout_rule. For an evaluation, evaluation_id is the evaluation of yours that opened it.
Read the subject
subject.display is masked to the first letter of each word, like A*** L***. subject.reference is your own reference for the subject, such as cust_7731, or null when you gave none. The API returns no name, date of birth or contact details at any point, so match on reference.
GET /v1/intelligence/reviews/{id} adds matched_entries, decisions, notes and events. Notes come back as presence only: who wrote one and when, never its text. Your team reads note text in the portal. A review that belongs to another account is a 404.
Deadlines
due_at is the time a decision is due, and overdue is true once it has passed without one. priority is high, medium or low. paused_at is set while the review is waiting_on_customer, and the due time stops running until you resume it. due_at can be null.
closed_at and outcome are null while a review is open. Your team's business hours and per-priority targets are in GET /v1/intelligence/reviews/settings, read only.
Events
Subscribe to intelligence.review.* with POST /v1/webhooks instead of polling.
| Event | When |
|---|---|
intelligence.review.opened | A review opened in your queue. |
intelligence.review.decision_proposed | A decision is waiting for a second approver. |
intelligence.review.decided | A decision took effect, including waiting on your customer. |
intelligence.review.reopened | A decision was sent back or withdrawn, or your customer responded. |
intelligence.review.closed | A review closed, with its outcome. |
intelligence.review.overdue | A review passed its due time without a decision. |
intelligence.review.note_added | A note was added. The note is never sent. |
Every event carries ids, codes and your own reference. It has no name, not even the masked one, and no date of birth, contact details, note text or person on your team. Fetch the review for more. reopened carries a reason: sent_back, withdrawn or customer_responded. Webhook Events has the payloads.
When it goes wrong
intelligence scope. A missing account capability is capability_required.status or kind, or a timestamp that is not RFC 3339 UTC with milliseconds and a Z.