agentmail is the OFAC sanctions screening layer for AI agents that transact autonomously. Built in 2025-2026 by a small team that watched the x402 and agent-wallet ecosystem grow faster than the compliance stack underneath it. The insight was simple: agents will increasingly sign transfers, and OFAC strict liability does not care whether the signer has a face or a wallet address. We built the infrastructure to close that gap before the first $330K fine hits the news.
We started with one question: what happens when an AI agent pays a sanctioned address? The answer, in a sentence: the operator does the time. OFAC penalties start at $330,944 per violation, strict liability applies, and there is no automation exemption. We dug into public data - the US Treasury SDN list, the vile/ofac-sdn-list multi-chain registry, and the patchwork of jurisdiction rules - and built a screening layer agents can call before they sign. It took longer than expected because edge cases in OFAC data are dense. We shipped when it was good enough to protect someone, not when it was perfect.
Every time your agent pays someone, agentmail screens the counterparty against the US Treasury OFAC Specially Designated Nationals (SDN) list before the transaction is signed. We check 782 sanctioned crypto wallets, 19,086 names, and 16 embargoed jurisdictions. If the counterparty is flagged, the transaction halts. If clean, it proceeds in under 100ms.
OFAC penalties start at $330,944 per violation. Strict liability applies even to automated transactions. As AI agents begin moving money autonomously, the gap between "cool demo" and "legally compliant" is a $330K fine. We close that gap with one line of code.
agentmail is MIT licensed and available on GitHub. You can self-host for free or use our hosted API with 5 free checks per day.
Questions? Open an issue on GitHub or email us. We respond to security and compliance questions within 24 hours.