Fraud Score API
Event signals in. Fraud probability, the reasons, and a decision out.

How we compare

The incumbents either make you train and host a model yourself (AWS Fraud Detector), gate the product behind an enterprise sales motion and a percentage-of-GMV contract (Sift, Kount, Signifyd, Riskified), or only score the payments that flow through one processor (Stripe Radar). Brainiall is a flat per-event REST call: send signals, get a scored decision with the reasons attached, improve it with your own labels — self-serve from the first call, no platform to adopt.

Prices are list-price approximations for orientation, not quotes — most fraud vendors price by negotiation. Always check each vendor's current terms.

Pricing

One per-event price for scoring; submitting outcomes via /feedback is unmetered. The free tier is enough to wire scoring into a checkout flow and start collecting feedback.

PAYG: $0.0075 / scored event (Brainiall Fraud engine). /feedback calls are free. No minimum spend, no contract, no percentage of your GMV.

Two calls: score, feedback

# 1. Score an event — send whatever signals you have; all fields are optional
POST https://api.brainiall.com/v1/fraud/score
  {"event_id": "ord_8821", "event_type": "payment",
   "amount": 4200, "currency": "USD", "avg_txn_amount_30d": 85,
   "txn_count_1h": 7, "distinct_cards_24h": 4, "distinct_countries_24h": 3,
   "is_new_device": true, "is_new_ip": true, "is_tor": true,
   "card_country": "US", "ip_country": "RU", "account_age_days": 0,
   "avs_match": false, "cvv_provided": false, "prior_chargebacks": 1}
  -> {"event_id": "ord_8821", "fraud_probability": 0.99, "risk_level": "critical",
      "decision": "deny", "risk_score_points": 28.0,
      "risk_factors": [
        {"factor": "amount_far_above_typical", "weight": 5.0, "direction": "increases_risk",
         "detail": "amount is 49.4x the 30-day average (4200.00 vs 85.00)"},
        {"factor": "tor_exit_node", "weight": 3.5, "direction": "increases_risk",
         "detail": "request originated from a Tor exit node"},
        ... ],
      "decision_bands": {"allow_below": 0.40, "review_below": 0.75},
      "model": "Brainiall Fraud engine"}
  # override the bands per request: "decision_bands": {"review": 0.30, "deny": 0.80}

# 2. Report the confirmed outcome so the model re-calibrates to your data
POST https://api.brainiall.com/v1/fraud/feedback
  {"event_id": "ord_8821", "label": "chargeback", "notes": "issuer dispute, fraud confirmed"}
  -> {"event_id": "ord_8821", "label": "chargeback", "accepted": true, "feedback_id": 42}
  # label ∈ fraud | legitimate | chargeback | dispute
GET https://api.brainiall.com/v1/fraud/feedback/stats
  -> {"total": 1240, "by_label": {"legitimate": 1180, "fraud": 41, "chargeback": 19}}

Every field is optional — start with the three or four signals you already have and add more over time; the score and the risk-factor list adjust accordingly. The response is fully explainable: risk_factors lists exactly which signals moved the score and by how much (positive = increases risk, negative = a mitigant), so you can show a human reviewer why an event was flagged.

What it's for

• Checkout / payment risk: score each transaction at authorization time and auto-approve the clearly-good, send the borderline to manual review, and block the clearly-bad — with the reasons attached for the review queue.
• Signup & account-creation abuse: catch fake-account farming and promo abuse from the velocity, device/IP novelty, disposable-email and VoIP-phone signals before the account does anything.
• Account-takeover signals: flag logins and account changes that come from a new device/IP, an anonymizing network, or a country that doesn't match the account's history.
• Card-testing detection: spot bursts of small transactions across many cards on one account — the classic stolen-card validation pattern.
• Tune it to your business: override the allow/review/deny probability bands per request, and feed confirmed outcomes (chargebacks, disputes, confirmed-good) back via /feedback so the model re-calibrates to your own fraud rate.
• One bill, one key: same Brainiall API key and usage-based billing as the rest of the catalog — no separate fraud vendor to procure, no percentage of GMV.

Press kit & resources

What reviewers, integrators and procurement teams typically ask for.

More specialty APIs

Same single API key, same usage-based pricing, different problem solved.

Identity VerificationFace detection, embedding & liveness

Voice IDCaller verification & account-takeover signals

Document IntelligenceID / invoice / receipt → structured fields

Content ModerationText & image policy screening