AI agents can move money, deploy code and call APIs. But logs only tell you what they claim happened afterward. Prove Before Act adds verifiable evidence before execution. The agent commits its declared WHY, anchors a cryptographic proof, acts, then records WHAT happened. The result is an independently verifiable sequence without exposing private data. Open standard. MCP, REST and Python/JS SDKs. 10 proofs free, then $0.01/proof. xProof is the reference implementation.