▣
Autonomous commerce
Reconcile invoice, payment, merchant, fulfillment, and receipt records without mistaking authorization for delivery.
Don't trust success.
Verify outcomes.
Autonomous systems can request actions, receive approvals, and report that the job is done. But a success message is not proof. OORTYEET VERIFY investigates what actually happened—against defined claims and the evidence available.
The public demo is a synthetic browser illustration; the private pilot has separately produced a Stripe TEST-mode draft. This site makes no live provider requests, payment transfers, or production-security claims.
OORTYEET / CLAIM ENGINELOCAL TEST
CASE 0082 · INVOICE CREATION
Was the requested invoice actually created correctly?
✓
RequestedInvoice $23.50 · Reference IN-082
CAPTURED✓
AuthorizedAgent permission & amount verified
ALLOWED✓
Execution reportedProvider acknowledged creation
OBSERVED!
Outcome reconciledDestination reports a different amount
CONFLICTBuilt for a harder question — not “did the agent say it worked?”
Why this exists
Four questions that traditional activity logs can confuse.
We make explicit claims about what is known, what is unsupported, and what would be needed to resolve uncertainty.
01 · INTENTION
Requested
Capture the exact action, actor, and requested outcome.
02 · PERMISSION
Authorized
Check delegated scope and deterministic rules before execution.
03 · PROVIDER
Executed
Record the destination's acknowledgment without assuming it is true completion.
04 · EVIDENCE
Verified
Compare destination readback and support only the claim the evidence permits.
The outcome problem
A signed message can be authentic and still be wrong.
Two messages from the same controller are not necessarily two independent witnesses. Our experimental evaluator tracks source control domains, excludes self-reports as outcome evidence, and preserves contradictions.
Inspect a live sandbox scenario ↗signed verification receipt · ExampleSCOPE LIMITED
{
"claim": "invoice_record_exists",
"request_amount": 2350,
"observed_amount": 9999,
"control_domain": "local_mock",
"decision": "CONFLICTING_EVIDENCE",
"does_not_prove": [
"payment", "delivery", "settlement"
]
}Where evidence matters
One evaluation architecture. Different standards of proof.
The commercial roadmap prioritizes invoice and software outcome verification. Physical-world claims require appropriately independent measurements and domain specialists.
⌘
Software delivery
Compare code-completion claims with CI results, deployment records, health checks, and security evidence.
⚙
Robotics & manufacturing
Eventually assess simulation-to-real gaps and physical outcomes using independent sensors and inspections.
◈
Scientific research
Link hypotheses, methods, instrument records, data, and reproducibility evidence to carefully scoped claims.
≋
Accounting & reconciliation
Identify inconsistencies across authorized actions, ledgers, processor records, and bank evidence.
⚡
Infrastructure & energy
Compare commanded changes with authoritative infrastructure and telemetry outcomes under safety controls.
Actual supported milestone
Try it. Trigger a disagreement. Inspect the receipt.
The interactive demo now calls the actual local verification engine through HTTP. Create a synthetic invoice, query the simulated destination separately, deliberately introduce a mismatch, and inspect the signed claim.
Runs only when the local sandbox backend is started; the public browser display never moves money.
Open verification laboratory →RECORDED OUTCOME
CONFLICTING_EVIDENCE
- Expected
- $23.50 USD
- Observed
- $99.99 USD
- Claim
- Invoice record matches
- Verdict
- Cannot certify completion
Early technical pilots
Have a workflow that says “done” before you know it's done?
We're evaluating real business problems where a verifiable destination record, a contradiction report, or better approval boundaries could reduce investigation costs.
Discuss a scoped pilot ↗What a pilot would measure
No claims of paying customers or independently audited production readiness.