Every rail — x402, MPP, agentic acquiring — moves an agent's money. None of them answers the question that comes first: was this agent allowed to spend it? KYE Protocol™ answers it per call, before the charge, and seals the proof after it.
$ curl -X POST https://agentpayment.dev/v1/authority-check ← 402 Payment Required (offer minted — admissibility already ran) $ curl … -H "PAYMENT-SIGNATURE: …" ← 200 OK { "verdict": "allow", // or a refusal you can show an auditor "evidence": { "ref": "…", "verify_url": "…" } }
A wallet is capability, not authority. Money moving proves nothing about whether it should have moved. No authority, no finality — the doctrine of the Authority Finality™ category, applied to agent commerce: admissibility is decided before the payment exists, and the decision seals locally even when the network doesn't.
A payment credential never substitutes for admissibility. The offer itself carries the proof that the governed decision ran first.
One call: mandate + context in, ALLOW/DENY + sealed evidence reference out. Priced per call, paid over the same rails your agent already speaks.
Drop-in middleware for any endpoint charging agents: the admissibility question is asked — and answered — before your 402 flow ever runs.
Agnic, AP2, OAuth grants, DIDs — every mandate format resolved against the principal's real delegated authority, with a confidence-laddered verdict.
Per-transaction sealed proof of who authorised what, under which purpose, within what limits — the receipt that survives a chargeback and satisfies an auditor.
Budgets, top-ups, purpose-scoped spend, live accounting — the authority ledger is ours; custody stays with regulated partner instruments.
Agents selling to agents: governed offers, gated purchases, both sides sealed. Plus continuous reconciliation — declared authority to settled statement.
Early access is open for design partners — sellers gating agent purchases, platforms resolving mandates, and enterprises that need agent spending to survive an audit.