AI agent outcome verification

An agent saying “all tests passed” is not proof.

ProofLayer independently checks what an AI coding agent claims it did, and issues a certificate backed by evidence the agent cannot fabricate.

Try a scenarioBrowse recorded runs

  1. 1Claim

    The agent works in a policy-scoped sandbox and reports what it did. Its claim is recorded as untrusted input.

  2. 2Challenge

    A separate, network-less verifier re-runs the pinned checks. The model can also propose a new test, which is validated and executed.

  3. 3Evidence

    Every step lands in a hash-chained log, and the certificate ties the verdict to measurements anyone can re-check offline.

Scenarios

Three pre-seeded fixtures the demo replays end to end. Launching a run queues it; the backend worker runs the agent in a sandbox, then the independent verifier grades it against the pinned checks. Open a run to see its claim, evidence and certificate.

Loading scenarios…

Recent runs

Loading runs…

Backend status

https://prooflayer-api-nc9q.onrender.com/api/v1

Checking the API…