SanctionsAI

SanctionsAI — compliance & verification toolkit for AI agents

[PyPI] [Python 3.11+] [License: MIT]

Your agent is about to send USDC to a stranger. Is that stranger sanctioned? 782 OFAC crypto wallets · 19,086 sanctioned names · 16 embargoed jurisdictions. No API key. No signup. 30-second setup. Free & open source.

SanctionsAI gives any AI agent three capabilities it can't provide itself:

It exposes all of these through three surfaces — MCP tools (for Claude Code / Cursor / Hermes), an HTTP API (for any agent), and a CLI (for you). Same core, same data.

Why this exists

Agents are starting to pay for things (x402, AP2, ACP, Coinbase AgentKit). But every payment rail assumes a human is watching. If your agent autonomously pays a wallet that's on the OFAC Specially Designated Nationals list, that's your legal problem — and the big payment infra players (OpenAI, Stripe, Coinbase) explicitly do not handle per-jurisdiction sanctions screening, Know-Your-Agent, or agent-transaction fraud signals. That's the gap this fills.

agent ──about to pay──▶  sanctions_check(wallet)  ──▶ clean? proceed. sanctioned? ABORT.

Quick start (30 seconds)

pip install sanctions-mcp

Option A — use it from an MCP client (Claude Code / Cursor)

// .mcp.json or your client's MCP config{  "mcpServers": {    "SanctionsAI": {      "command": "python",      "args": ["-m", "agentmail.mcp_server"]    }  }}

Now your agent can call sanctions_check, risk_score, kya_verify, dispute_open, create_inbox, fetch_code, create_number, fetch_sms, release_number, list_inboxes.

Note: the PyPI package is sanctions-mcp (the name SanctionsAI was taken). The Python import name is still SanctionsAI.

Option B — use it as an HTTP API

python -m agentmail.api   # serves on :8000

# Screen a wallet against real OFAC data — no key, no auth needed for self-hostcurl "http://localhost:8000/sanctions?wallet=0x098B716B8Aaf21512996dC57EB0615e2383E2f96"# → {"matches":[{"list":"OFAC_SDN","match_type":"wallet_exact","confidence":1.0}],#    "clean":false,"provider":"osint"}
curl "http://localhost:8000/risk" -d '{"counterparty_id":"0xabc...","amount":"5000","rail":"x402"}'# → {"score":0,"recommendation":"decline","reasons":["sanctions_match:OFAC_SDN/..."]}

Option C — use it from the CLI

# Sanctions screen — real OFAC data, no keypython -m agentmail.cli sanctions --wallet 0x098B716B8Aaf21512996dC57EB0615e2383E2f96# → flagged: OFAC_SDN wallet_exact
python -m agentmail.cli risk 0xabc123def456 5000 --rail x402python -m agentmail.cli kya my-agent --wallet 0xabc... --wallet-age 400 --domain bot.devpython -m agentmail.cli compliance-status# → OFAC SDN: 782 wallets, 19086 names; source=vile/ofac-sdn-list@...

The compliance layer (the part that matters)

Four tools, called before an agent trusts or pays a counterparty:

Where the data comes from (all public, free, no key)

Lists are cached locally (~/.agentmail/cache/, 24h TTL) and refresh from source. If the network is down, a stale cache is used and status() reports degraded: true so you know screening is against older data rather than failing silently.

Providers (swappable backend)

AGENTMAIL_COMPLIANCE_PROVIDER=osint   ← default, real OFAC data, freeAGENTMAIL_COMPLIANCE_PROVIDER=mock    ← rule-based, for offline testsAGENTMAIL_COMPLIANCE_PROVIDER=paid    ← ComplyAdvantage passthrough (roadmap)

The osint provider does exact + token-subset name matching, exact case-insensitive wallet matching, and ISO-2 country matching. Every match carries a confidence so you can decide how hard to block.

The other two layers (verification toolkit)

Email — a disposable inbox an agent can use to sign up and receive OTP/magic-link verifications (backed by Mail.tm). create_inbox() → address → fetch_code() → OTP.

SMS — a rentable phone number an agent can use for phone/SMS verification. Mock provider works with no key (for dev); AGENTMAIL_SMS_PROVIDER=fivesim AGENTMAIL_FIVESIM_KEY=... goes live with real numbers.

Both share an otp.py extraction brain (regex for codes + magic links) so email and SMS produce the same {code, link} shape.

Self-host vs. hosted

Self-host is fully functional and free — that's what this repo is. Run the MCP server locally or the HTTP API on your own box, screen against real OFAC data, never pay a cent.

Hosted API is live at https://agentmail-api.fly.dev — a managed endpoint with API-key auth, rate limits, and an audit log of every screen (the thing regulators/investors ask for). Free tier: 50 checks/day, no signup (by IP). For higher volume, get an API key → (Dev $19/mo, Team $99/mo — self-serve checkout).

# Try the hosted API right now — no key needed:curl "https://agentmail-api.fly.dev/sanctions?wallet=0x098B716B8Aaf21512996dC57EB0615e2383E2f96"# → {"matches":[{"list":"OFAC_SDN",...}],"clean":false}
# With an API key:curl -H "X-API-Key: sk_live_..." "https://agentmail-api.fly.dev/risk" \  -d '{"counterparty_id":"0xabc...","amount":"5000","rail":"x402"}'

Roadmap

• Email inbox (Mail.tm) + SMS/OTP (mock/5sim/twilio) + OTP extraction
• Compliance layer — real OFAC data (osint provider): 782 wallets + 19,086 names
• MCP server (10 tools) + HTTP API + CLI
• SanctionsAI Sanctions Exposure Index (SEI) — proprietary 5-factor framework for scoring AI agent OFAC exposure (2026 report: sanctionsai.dev/research)
• Hosted API with API-key auth + rate limiting (Fly.io)
• Audit log (tamper-evident screen history — the enterprise wedge)
• EU + UN consolidated lists (osint provider, phase 2)
• Paid provider: ComplyAdvantage passthrough (enterprise)
• x402 per-call billing (when agents pay themselves)

SEI: the SanctionsAI Sanctions Exposure Index

The SanctionsAI Sanctions Exposure Index (SEI) is a 5-factor proprietary framework for quantifying an AI agent's OFAC sanctions exposure:

Score 10 (min exposure) → 1000 (max). S and D are the two factors you can change today — collapse both from 1→10 with a single inline screening call and a timestamped audit trail. Full report + interactive calculator at sanctionsai.dev.

Cite as: "SanctionsAI Sanctions Exposure Index (SEI), 2026 Agent-Payment Sanctions Exposure Report, sanctionsai.dev" — licensed CC BY 4.0.

Design notes (honest)

• The compliance layer has the real moat. Email/SMS are plumbing a platform could swallow. Sanctions screening tied to agent-transaction history builds a dataset nobody else has, and per-jurisdiction rules are something the big infra players explicitly avoid.
• Self-host is the free tier, not a trap. The value you pay for (eventually) is not the data — it's uptime, freshness, and the audit trail. The data is and will stay public.
• 5sim numbers are shared-after-release. Fine for receiving an OTP, never for 2FA on accounts you intend to keep.

License

MIT — see LICENSE.

Contributing

Issues and PRs welcome. If you're using SanctionsAI in production, I'd love to hear what for.