Agent-verified work settlement,
powered by Somnia agents
Submit evidence on-chain. Agents screen it against requirements. A bounded verdict drives settlement policy, while ambiguous cases remain reviewable.
Agents screen evidence. Judges and sponsors select finalists. Contracts enforce payouts.
Two products, one protocol
Fund builders at scale.
Agents screen every applicant.
Sponsor locks a prize pool, sets on-chain criteria, and opens applications. Each applicant submits a public evidence URL — agents fetch it, extract facts, and return a verdict. You pick finalists from a pre-screened shortlist.
Commit to deliverables.
Escrow releases on verified proof.
Client defines requirements and locks payment. Contractor delivers, posts evidence, and agents verify autonomously. Complete verdict makes payment claimable — claim policy set at creation.
How it works
Three-agent pipeline
JSON API Agent fetches evidence · LLM Parse Agent reads project context · LLM Inference Agent returns verdict. Each step is auditable
Bounded verdicts only
Agents can only return Complete, Needs Review, or Incomplete — never move funds. Settlement logic lives entirely in the smart contract
Immutable proof trail
Every submission, agent request, verdict, and payout is recorded on-chain. Any party can verify the full history at any time