A correspondent bank lets a counterparty's agent settle — without trusting its API
The situation
A regional bank receives payment intents from a correspondent's autonomous settlement agent. Compliance cannot approve software that acts on an unverifiable claim of authority — and cannot call the correspondent's own API to check, because that API is the very thing under question.
What PRAXIS changes
Registration and attestation become on-chain facts before a single intent is submitted. At settlement, the receiving system re-checks reliance itself — it never trusts a cached result. A payment outside scope, or under a revoked CPoA, structurally cannot settle.
The outcome
The compliance desk relies on a record it can re-derive independently — active, unrevoked, in-window, signature-verified at the instant of settlement — and a transparency-log receipt it can produce point-in-time, months later, for an examiner.
Illustrative scenario built from the protocol's interbank-settlement reference flow. Not a claim about a specific named institution.
The control flow
Every flow reaches settlement through an unbroken chain of checks — or terminates in a structured denial. Never a silent success.