Live on Cloudflare Workers · Free while in beta

Screen every AI agent payment before it settles.

OFAC sanctions screening for autonomous payments. One API call. Verdict in under 50ms. Full audit trail. No signup required — get a key instantly.

Get Free API Key → Try Live Demo

<50ms

Verdict latency

38K+

SDN entities

650+

Sanctioned wallets

Chains

How It Works

One call between your agent and settlement.

ClearAgent screens the counterparty before money moves. Register once, screen every payment, get a durable audit trail.

Your agent

Initiates payment

→

ClearAgent

Screen <50ms

OFAC · PEP · Policy

→

Settlement

Payment clears

Clear

Clean — payment settles

OFAC clear, policy within limits. Audit trail logged automatically.

Review

Held for human approval

PEP match or adverse media flag. BSA officer reviews before settlement.

Block

Stopped pre-settlement

Sanctions hit or policy violation. Payment never clears. Operator notified.

request
# Screen before payment
curl -X POST https://api.clearagent.dev/v1/screen \
  -H "Authorization: Bearer <token>" \
  -d '{
  "token":   "<agent-jwt>",
  "cpName":  "Acme Corp",
  "cpWallet": "0x71C7656EC7ab88b...",
  "amount":   50000,
  "currency": "USDC",
  "chain":    "ethereum"
}'

           200 OK
          23ms
        
{
  "txnId":          "txn_8Kx2mQ9v...",
  "verdict":        "CLEAR",
  "riskScore":      2,
  "responseTimeMs": 23,
  "signals": {
    "ofac":        "CLEAR",
    "pep":         "CLEAR",
    "spendPolicy": "CLEAR"
  },
  "screenedAt":    "2026-03-21T14:32:18Z"
}

Integrates with Stripe MPP x402 ATXP Fiat rails Raw address

Under the Hood

Not a wrapper. A purpose-built screening engine.

ClearAgent runs on Cloudflare Workers at the edge — no cold starts, no containers, no third-party proxy. Everything executes in the same V8 isolate that handles the HTTP request.

38K+

SDN entities matched in-memory

Full OFAC SDN list parsed from CSV nightly via GitHub Actions. Stored in KV, loaded once per isolate. Exact match, substring match, and fuzzy Levenshtein matching with Unicode homoglyph defense — catches Cyrillic evasion attempts like "GАRАNTEX".

deterministic screening rules

Weighted rule engine with explicit precedence. OFAC hits (weight 35–40) override everything. Credential violations (25–45) next. Spend policy checks (8–15) last. No ML black box — every verdict is explainable and auditable.

token auth model

Operator gets a 1-year KYAOperatorCredential. Each agent gets a 90-day KYAAgentCredential (ES256 JWT) with embedded spend limits, allowed chains, and mode (supervised vs. autonomous). Credentials are the policy — no config files, no dashboards to forget.

screening log for audit trail

Every screen writes to Cloudflare D1 — txn_id, verdict, risk score, rule triggered, response time. REVIEW verdicts enter a resolution queue. Operators can pull their full audit history via API. Built for FinCEN exam day, not just demo day.