Score oversight in one request.
A 50-endpoint scoring API behind a documented judgment envelope, plus dedicated regulator-report, signed-receipt, evidence-chain, and judgment-velocity services. Post the telemetry for one human-AI interaction; receive a Judgment Independence Score across seven dimensions and a Decision Risk Score route. Signed receipts are issued from the DRS gate path.
A score in one call.
Authenticate with a scoped API key in the X-API-Key header and post the telemetry for one human-AI interaction. The response is synchronous and documented.
DRS routes the decision, JIS scores the router.
Use both together. The Decision Risk Score routes each AI-assisted decision; the Judgment Independence Score measures whether the reviewer is a trustworthy router.
POST /v1/decision/score
Submit one AI-assisted decision for Decision Risk Score routing. Returns the decision risk score, the routing decision, policy trigger evidence, and an audit identifier.
POST /v1/score
Score the human-AI interaction telemetry and return the Judgment Independence envelope across the seven dimensions, the band, and the decay projection.
Designed to be cited, not just called.
Judgment envelopes
Score responses carry documented judgment payloads and request metadata. Signed receipts are issued from the DRS gate path and receipt service.
API key authentication
Scoped API keys per environment, sent in the X-API-Key header. Rotate without downtime. Keys never appear in evidence receipts.
Synchronous scoring
Scores return in band on the same request. No polling, no webhooks required for the common path.
Typed SDKs
First party clients for TypeScript and Python, generated from the same schema as the docs.
Mapped to the spec
Each dimension in the response maps to a clause in the specification and a regulatory citation.
Observable
Latency and uptime are published. Incidents are posted to the status page with a timeline.
The evidence a regulator, an auditor, or a buyer asks for.
The scoring API returns the number. These services turn that live evidence into the artifacts other people need: a signed report, an exportable receipt, an append-only reasoning chain, and an early-warning velocity signal. Each runs on its own dedicated subdomain and reuses your existing API key.
Regulator report
A signed compliance report that maps your live oversight evidence to US and EU AI oversight law and NIST AI RMF 1.0. Generated on demand, ready to hand to a regulator or an auditor.
DDQ kit
Answers thirteen AI-oversight question categories from standard vendor security questionnaires (CAIQ, SIG, VSQ) using your measured behaviour, not asserted policy. Speeds procurement review.
Signed evidence receipt
An exportable signed per-decision evidence artifact serialized from an existing decision record. Anyone with the receipt can check whether the contents changed after COHESION issued it.
Append-only evidence chain
Capture a reviewer's reasoning before the AI answers, then bind it into a per-org HMAC hash chain. The reveal-gate keeps the AI output sealed until the human commits, so the record cannot be back-fit.
Judgment velocity
A predictive slope over a reviewer's recent Judgment Independence Scores, so a reviewer drifting toward rubber-stamping is caught before the score crosses the oversight floor, not after.
Receipt checker
A customer, auditor, or buyer can paste an exportable receipt into the checker and see whether the signature still matches the receipt contents. The checker does not reveal receipts the person does not already have.
Selected endpoints.
A sample of the scoring API surface. The full reference, schema, and error catalog live in the docs.