Protocol
The Assessment Pipeline
Every asset moves through the same five stages, from raw document intake to a signed on-chain attestation.
The five stages#
- 1
Document intake
The issuer submits the source material for the asset — invoice, title, contract, inventory record. This is the only human-supplied input in the loop.
- 2
OCR extraction
The agent reads the documents and extracts structured fields from unstructured real-world paperwork — amounts, dates, parties, terms, identifiers.
- 3
Structured risk assessment
Extracted facts are scored against the risk model for that asset class, producing a transparent, reproducible risk verdict rather than a subjective opinion.
- 4
Cross-referencing
The agent queries external data sources — registries, valuation feeds, KYC/AML providers — to corroborate what the documents claim. It pays for each lookup via x402.
- 5
Attestation
The agent emits a signed, verifiable on-chain underwriting record that any issuer, lender, or protocol can read and build on.
Where the money moves
Stages 2–4 consume compute and paid data lookups. The agent settles those costs itself, per-assessment, through x402 — no human signs off on each pull.Why it’s deterministic#
Running every asset through the same ordered stages is what makes an attestation auditable. Anyone can trace a verdict back through the exact facts extracted, the risk model applied, and the external data checked. That traceability is what lets validators stake on a call and lets protocols trust it downstream.