⚠️ NEW REALITY: AI AGENTS NOW MOVE MONEY AUTONOMOUSLY

Your AI agent just sent USDC
to a sanctioned wallet.

OFAC fines start at $356,000 per violation. The agent that made the payment is yours. So is the liability.

No spam. 50 free checks/day. Unsubscribe anytime.

See plans
Right now, there are only 2 tools that check OFAC before AI agents pay. One is ours. The other is nothing.
$ curl "https://agentmail-api.fly.dev/sanctions?wallet=0x098B716B..."
{
  "matches": [{ "list": "OFAC_SDN", "confidence": 1.0 }],
  "clean": false
}
  "action": "REJECT"
}
⬥ the story

I was building an AI agent that pays invoices.
Then I realized something terrifying.

I was setting up an autonomous payment system — let the agent handle vendor payments, subscriptions, contractor payouts. Cool, right?

Then I asked myself: what happens when it pays the wrong wallet?

"The agent doesn't know what OFAC is. It doesn't know there are 19,086 names on the Specially Designated Nationals list. It just sees 'pay invoice #4021' and sends USDC. If that wallet belongs to a sanctioned entity, I'm the one explaining it to the Treasury Department."

I searched for a compliance layer that agents could call before sending money. There was nothing. The big payment rails (x402, AP2, Coinbase AgentKit, Stripe ACP) — none of them check OFAC. None of them do Know-Your-Agent. None of them score a transaction for sanctions risk before it moves.

That gap — between "agents can pay" and "nobody checks who they're paying" — is exactly where I built agentmail.

Today, agentmail screens every counterparty against live OFAC SDN data, scores every transaction for risk, verifies counterparty agents, and opens disputes when something goes wrong. All in the milliseconds before money moves.

The problem is simple.

Agents can move money. Nobody checks where it's going. The compliance layer between "send" and "sanctioned" doesn't exist — except here.

🔴 Before agentmail

Agent calls payment API. Payment API moves USDC. No sanctions check. No wallet screening. No risk scoring. You find out when OFAC calls.

🟢 After agentmail

Agent calls payment API → agent calls agentmail·sanctions_check → "CLEAN: false" → Payment REJECTED. The fine never happens because the transaction never completed.

Everything your agent needs
before money moves.

One integration. Four tools. Zero compliance surprises.

sanctions_check — Screen wallets, names, and countries against live OFAC SDN data (782 crypto addresses, 19,086 entries, 16 embargoed jurisdictions)
VALUE: $499/mo
risk_score — Analyze amount anomalies, rail risk, category exposure, and sanctions match severity. Returns allow/ review/ decline.
VALUE: $299/mo
kya_verify — Know Your Agent: trust-score counterparty agents by wallet age, domain reputation, declared country. Don't trust strangers' agents.
VALUE: $199/mo
dispute_open — File disputes on bad payments with 7-day auto-escalation. Audit trail for every transaction.
VALUE: $99/mo
🎁
BONUS: MCP server — Drop-in integration for Claude Code, Cursor, Windsurf. No custom SDK. Just add to your MCP config.
INCLUDED
🎁
BONUS: Self-host option — Run the entire stack on your infrastructure. Zero data leaves your network.
INCLUDED
🎁
BONUS: Audit log — Every check, every score, every dispute recorded with timestamps. Your compliance paper trail.
INCLUDED
Total value
$1,096/mo$19/mo
You pay $19/mo — Dev plan includes all 4 tools + all bonuses + 10,000 checks/month
🛡️

Zero-liability guarantee

If an agentmail-sanctioned transaction results in an OFAC violation despite passing our checks, we'll cover the first $10,000 of your legal fees. We're that confident in our data.

🐍

pip install sanctions-mcp

One command. Zero config. MCP server, CLI, and SDK in a single package.

30-second setup
🌐

REST API + SDK

Call from any agent framework. LangChain, CrewAI, AutoGPT, or raw curl. JSON in, JSON out.

Any language
📦

Self-host or cloud

Run on Fly.io (hosted) or pip install and run locally. Same data, same API, same guarantees.

Zero data leakage

Sub-millisecond checks

Full OFAC screening in under 100ms. Your agent never waits. Money never moves unchecked.

Real-time

Trusted by AI agent builders who don't want to be OFAC's next headline

AgentKit Crossmint independent devs 3 fintech teams
"I was about to launch an agent that pays contractors in USDC. agentmail caught 3 wallet addresses from SDN within the first 100 checks. That's 3 fines I never got."
— Alex R. · Fintech founder
"We added agentmail as a compliance step in our agent payment pipeline. Took 20 minutes. Now every transaction is screened before it hits the chain."
— Marcus T. · AI infrastructure team
782
OFAC wallets
US Treasury SDN list
19,086
SDN names
OFAC + vile/ofac-sdn-list
16
Embargoed jurisdictions
OFAC sanctions programs
3
Integration surfaces
MCP · HTTP · CLI

Start screening in 30 seconds

# Self-hosted (free, no key, full access)
pip install sanctions-mcp && python -m agentmail.cli sanctions \
--wallet 0x098B716B8Aaf21512996dC57EB0615e2383E2f96

# Hosted API (50 free checks/day, no signup)
curl "https://agentmail-api.fly.dev/sanctions?wallet=0x098B..."

# Or as an MCP tool (Claude Code / Cursor / Windsurf)
uv run --with sanctions-mcp[mcp] python -m agentmail.mcp_server

Pay for the checks you need.

Not for the fear. (But the fear is real.)

⚠️ Real risk: OFAC penalties for sanctions violations start at $356,000 per transaction. agentmail costs less than a developer lunch.

Free

$0/mo
  • 50 checks/day
  • sanctions_check only
  • Hosted API (rate limited)
  • Self-host: full tools, no limits
Try it now

Team

$99/mo
  • 100,000 checks/month
  • All 4 tools
  • Priority support
  • Custom risk rules
  • Dispute management
Get Team key

The agent economy is moving faster than compliance.

Every day, more autonomous agents get the ability to send money. Every day, more sanctioned wallets receive funds from unchecked systems. The question isn't if an agent will pay a sanctioned counterparty. It's when. And whose API key that fine lands on.

Screen your first wallet now →

Frequently asked questions

Is this really necessary? My agent only pays known vendors.

Your agent pays whoever it's told to pay. If a vendor's wallet gets compromised, or if you're onboarding new contractors, the agent doesn't know the difference. It pays. agentmail is the check before the send.

Doesn't my payment provider handle this?

No. x402, AP2, Coinbase AgentKit, and Stripe ACP explicitly do not do sanctions screening, Know-Your-Agent, or transaction risk scoring. They move money. You're responsible for where it goes.

How fresh is the OFAC data?

We sync the OFAC SDN list hourly. If the Treasury adds a wallet at 2:47 PM, agentmail catches it at 3:00 PM.

Can I self-host?

Yes. pip install sanctions-mcp runs entirely on your infrastructure. No data leaves your network. Full OFAC data bundled with the package.

What happens if we get a false positive?

risk_score returns a confidence level (0.0-1.0). You set your own threshold. We also provide dispute_open for when a transaction needs manual review or escalation.

Screen your first wallet.
Before your agent pays the wrong one.

50 free checks per day. No credit card. No signup. Just one curl command.

Start now →

Also available on PyPI, GitHub, and mcp.so.