MoltProof Verifies Whether Autonomous Trading Agents Keep Their Word
MoltProof is a read-only verifier that proves whether on-chain autonomous trading agents actually followed the rules they publicly committed to, checkable by anyone.
Autonomous agents are increasingly trading real capital on-chain, whether their own, delegated funds, or DAO treasury assets. While blockchains transparently record what an agent did, they never record what it was authorized to do — that mandate typically exists only as an off-chain screenshot or claim. Copy-trading data exposes how costly this gap is: 97% of lead traders are profitable, yet only 43.6% of their copiers actually make money.
MoltProof is a read-only verification layer built to close this gap. Before trading, an agent publishes a signed, public mandate specifying allowed venues, output tokens, position caps, and a validity window, tied to its DID/ERC-8004 identity. MoltProof then pulls the agent's actual on-chain transactions and checks each one against that commitment, issuing a verdict — ADHERENT, BREACHED, NO_MANDATE, or NEEDS_REVIEW — and citing the exact transaction behind any breach.
Verdicts are Ed25519-signed, but the real trust comes from the fact that anyone can recompute the same result from public chain data and the public mandate. MoltProof never touches private keys, funds, or transaction signing, so even a compromised endpoint poses no financial risk — it can only return a falsifiable wrong answer. A live demonstrator on Base mainnet shows an agent restricted to buying WETH getting flagged after one swap into USDC, verifiable with a single API call. The system supports Base, Ethereum, Arbitrum, Optimism, Polygon, Hyperliquid, and Solana (beta), with API and MCP integrations and micropayments handled via x402.