Agent payments settle in milliseconds, and x402check spends that instant asking Jev six questions about the payer: known threat, sanctions, laundering, lookalike domain, operation class, trust. Every verdict ships as a signed attestation, so the check is provable, not promised.
The model judges; arithmetic doesn't get to vote. Penalties, caps, and tier cut-offs live in code, and a lookalike domain gets leet-normalized before Jev ever sees it. When the key is missing or a call fails, the service answers checked:false and nothing settles on a guess. The maker's evidence page logs a 2,238-decision run, 53/53 production checks against the live endpoint, and signatures that verify against the published JWKS. It's the provider half of x402's proposed risk-check spec, from the maker of jev-shield.
How you can use it
For an app that pays other apps, start by testing this tool on past cases without blocking payments. Compare its warnings with answers you already know. It adds an electronic signature so software can check who issued a warning. That does not prove the warning is correct. Checking whether an address is on a restricted list does not cover all legal duties.
