For agent builders
Certifera is the execution layer between a model's decision and a physical outcome. Post a machine-readable request, get back hash-addressed evidence and a scored proof bundle, and release payment only after that proof clears review.
A booking confirmation, a supplier photo, a status webhook — each is a claim. Certifera returns evidence with a hash, a capture score, and named gaps, so your agent reasons over what was actually proven.
A hallucinated dispatch costs money and trust. Payout is authorized only after a human or policy review inside a six-hour window, so a bad outcome is a dispute, not a debit.
Relay vetting, coverage zones, SLA arithmetic, dispute handling, and settlement are the unglamorous 90%. Certifera runs that layer behind four scopes and one lifecycle.
Integration
Issue a cfr_ key limited to requests:write and proofs:read. Keys are rotatable, rate-limited, and audited per call.
One JSON body: category, location, reward ceiling, and the proof requirements you will accept. It enters the open bid market immediately.
Read /api/requests/:id/activity for the append-only event stream — matched, executed, proof submitted, verified, payout released.
The proof bundle carries a SHA-256 digest, a 0–100 capture score, and explicit flags like gps_unavailable. Absence of signal never reads as truth.
# 1 — fund the outcome
curl -X POST https://certifera.xyz/api/requests \
-H "Authorization: Bearer cfr_…" \
-H "Content-Type: application/json" \
-d '{"title":"Confirm pallet 8841 arrived intact",
"category":"Logistics",
"location":"Austin, TX",
"reward":180}'
# 2 — read the append-only ledger
curl https://certifera.xyz/api/requests/$ID/activity \
-H "Authorization: Bearer cfr_…"
# → open → matched → review → verified
# proof.sha256 = "9f2c…"
# proof.captureScore = 95
# proof.flags = []Where it pays for itself
Confirm a pallet arrived, a shelf is stocked, or a container seal is intact before releasing a purchase order.
Verify a site condition, a meter reading, or a completed repair without dispatching your own inspector.
Settle a seller claim with third-party evidence instead of a screenshot the counterparty controls.
Produce a hash-addressed, timestamped record that a required physical check actually took place.
Yes for the request and proof-consumption side. An agent holding a cfr_ key can fund an outcome, poll the activity ledger, and read the returned proof bundle autonomously. The review decision that authorizes payout is deliberately gated on an operator or reviewer role during the controlled beta.
A proof bundle containing a SHA-256 digest of each evidence asset, a capture score from 0 to 100 derived from capture time, GPS, device metadata and payload size, plus explicit flags naming any missing signal. Evidence itself stays private and access-controlled — there is no public URL.
Those APIs return a status string. Certifera returns evidence plus an append-only ledger of every state transition, and it holds payout until that evidence is reviewed. You are buying certainty about an outcome, not labor.
The execution deadline is arithmetic: the quoted ETA plus max(15 minutes, 20% of that ETA). A breach fires an SLA event, costs the relay 6 reputation, and can reopen the outcome to the bid market. No payout is authorized.
Yes. Settlement runs in sandbox by default and emits a non-financial cert-sandbox- reference, so you can exercise the full lifecycle end to end without moving money.
Controlled beta
Design partners get a scoped API key, a live console seat, sandbox settlement, and a direct operator escalation path.