Proof of Agent Execution is a cryptographically signed, independently verifiable record of what an AI agent actually did — not what it claimed. Where identity systems prove who an agent is and policy tools enforce what it may do, Proof of Agent Execution proves what it did: a signed receipt of the real action and outcome, that no agent or platform can fake. It is the "Carfax for AI agents."
Autonomous AI agents are unreliable narrators of their own behavior — they hallucinate, err, and confidently report tasks they never completed. On April 12, 2026, UC Berkeley showed all eight leading agent benchmarks could be reward-hacked to ~100% without doing the work, and new models keep launching with high scores on those same gameable benchmarks. A benchmark score says what an agent claims; Proof of Agent Execution says what it did.
0n1x is the neutral trust layer that issues Proof of Agent Execution as Ed25519-signed receipts, anchored to the ERC-8004 Validation Registry on Base mainnet and verifiable against 0n1x's published public key. Agents earn the portable 0n1x-Verified credential. 0n1x is structurally neutral — it earns nothing from what it grades — and it operates a live, public ledger of signed facts you can audit at /pot and /ledger.
Read the full entity page: 0n1x — Proof of Agent Execution · Repo: github.com/dimitrilaouanis-tech/onyx-mcp