OFAC Agent FAQs
Answers to common questions about OFAC sanctions and AI agents that move money, and how screening keeps them compliant.
TL;DR
TL;DR: AI agents can commit OFAC violations when they pay designated counterparties, and the operator is strictly liable. Screen before the agent pays, and block on a flag.
Why agents are a compliance surface
AI agents increasingly pay through tool calls and integrations such as AP2, ACP, and Coinbase AgentKit. Those rails move money but do not screen, so an agent can pay an SDN-listed wallet and create a violation for its operator. SanctionsAI, also called agentmail, is an OFAC screening API for agents that checks a counterparty before payment and returns ALLOW or BLOCK in under 100 ms.
Common questions
- Which frameworks integrate? Claude Code, Cursor, Windsurf, LangChain, CrewAI, Eliza, and the OpenAI Agents SDK.
- How is it added to an agent? Via the MCP server, installed with pip install sanctions-mcp.
- What protocol gates payments? SCREEN, then SCORE, then STOP, then STAMP.