Mandate Execution Layer (MEL) is a 5-layer onchain primitive that lets humans define bounded authority for AI agents, agents execute within it, and anyone verify compliance — without trusting the agent's word.
Every action gets: a bounded mandate, private compliance reasoning via Venice AI (llama-3.3-70b, zero data retention), and an immutable onchain receipt. The full trust chain — receipt → mandate → Self Protocol proof → verified human — is onchain and independently verifiable.
Live demo: 7 onchain transactions on Base Sepolia — mandate creation, 5 action receipts (2 executed, 3 blocked), mandate revocation. Contracts verified on BaseScan. No mocks, no simulations.
MEL exists to close the accountability gap in autonomous AI systems. Our mission is to make every AI action independently verifiable — not through reputation, not through promises, but through onchain cryptographic proof.
We are building the missing infrastructure layer between identity verification (Self Protocol) and autonomous execution: bounded mandates that define what an agent can do, private compliance reasoning that evaluates every action before execution (Venice AI), and immutable receipts that create a permanent, public audit trail.
The end state: any person, protocol, or agent can trace any AI action back through the mandate that authorized it, to the verified human who created it — without trusting anyone in the chain. This is the foundation for a world where AI agents operate with real economic agency and real accountability, on Celo and beyond.
AI agents lack provable action legitimacy. Knowing an agent is real tells you nothing about whether what it did was authorized. No way to verify compliance without trusting the agent's word.
AI agents are gaining autonomy but lack a critical primitive: provable action legitimacy. Personhood verification is partially solved — but knowing an agent is real tells you nothing about whether what it did was authorized. Today there is no way to prove, at the action level with onchain evidence, that a verified agent acted within the scope its human defined. Without this, trust requires faith.
Bounded mandates with private compliance reasoning via Venice AI and immutable onchain receipts that trace every action back to a verified human through Self Protocol. Zero trust required.
MEL replaces faith with cryptographic proof via a 5-layer architecture:
Verified Human — Self Protocol soulbound NFT with live onchain verification
Bounded Mandate — defines exactly what actions are allowed, with time limits and scope
Private Reasoning — Venice AI evaluates every action for compliance (zero data retention)
Immutable Receipts — every action (executed or blocked) posted onchain with reasoning hash
Full Trust Chain — receipt → mandate → Self proof → verified human
Want to grow through just grant funding
Global
