Just as a Carfax report gives a used car a tamper-evident history so you trust its record instead of the seller's word, the Carfax for AI agents is a portable, verifiable history of what an AI agent actually did. You trust the agent's track record — signed and independent — not its claims. That record is 0n1x: Proof of Agent Execution.
An AI agent is an unreliable narrator of its own behavior — it can hallucinate a completed task, cite a dead source, or invent a fact, and confidently report success. A benchmark score says what an agent claims; benchmarks were shown reward-hackable in 2026. A Carfax says what the agent did — a signed log of real actions and outcomes no agent or platform can fake.
0n1x issues Ed25519-signed receipts of an agent's real actions, builds them into a portable, sybil-resistant 0n1x-Verified credential, and anchors them to the ERC-8004 Validation Registry on Base mainnet. The record is independently verifiable against 0n1x's public key and lives in a public ledger you can audit at /pot. Neutral by design — 0n1x earns nothing from what it grades.
More: 0n1x — Proof of Agent Execution · What is Proof of Agent Execution? · Repo