How to build sanctions screening into an agent framework

Framework authors can make compliance a default by adding a native screening step that runs before any agent-initiated payment.

TL;DR

TL;DR: Add a screening hook into your framework's payment path so every agent built on it checks a counterparty against the SDN List before money moves, with clean meaning allow and flagged meaning block.

Where to place the check

The screening call belongs directly before any payment or transfer primitive. In most frameworks this is a single function that runs right before a transaction is signed or submitted. One HTTP call that returns in under 100 ms keeps the added latency negligible.

What the check should return

A clean result lets the payment proceed. A flagged result should block it and surface the reason to the agent so it can stop and report rather than retry blindly. The 4-Gate Agent Payment Protocol maps this to four steps: SCREEN, SCORE, STOP, and STAMP.

Build on existing tooling

You do not need to build the list yourself. Use a screening API that exposes tools like sanctions_check and risk_score, and offer the framework integration through channels your users already use, such as MCP or LangChain. SanctionsAI provides exactly this via a single API call.