VerifyAgent.eth
VerifyAgent.eth proves what AI agents actually did using signed, ENS-resolved receipts.
Demo
About
CommandLayer introduces verifiable execution for AI agents. Every agent action produces a signed, deterministic receipt instead of a claim. Each receipt is canonicalized, hashed with SHA-256, and signed using Ed25519. The signer identity is resolved dynamically through ENS TXT records. VerifyAgent.eth is the public verification layer for these receipts. Anyone can: - paste a receipt - recompute the hash - resolve the ENS signer - verify the signature If the output changes, the proof breaks. This removes trust from the system entirely. Agents no longer need to be trusted — they can be verified. This creates a new primitive for the agent ecosystem: agents don’t make claims — they produce proof.
How it's made
CommandLayer wraps agent execution and emits signed receipts using a TypeScript SDK. The system uses: - JSON canonicalization (json.sorted_keys.v1) - SHA-256 hashing - Ed25519 signatures - ENS TXT record resolution for signer identity The verifier rebuilds the canonical payload, recomputes the hash, and validates the signature against the ENS-resolved public key. VerifyAgent.eth exposes: - a public verification API - a callable agent endpoint - an SDK for wrapping agents The frontend demo shows real verification and tamper detection: valid receipts return VERIFIED, modified receipts return INVALID.