UCP-Compatible Agent-Authorized Payments via Closed-Loop Wallet Settlement

<div> This paper describes a model for executing the payment leg of an agent-driven commerce flow as a wallet-to-wallet transfer within a single closed-loop wallet operator. The model is positioned as a wallet-native settlement implementation that can sit behind a checkout/payment step in a Universal Commerce Protocol (UCP) commerce journey; it is not a replacement for UCP. UCP governs the broader interaction between agents, merchants, and commerce systems - discovery, catalog, cart, checkout, order state, and post-purchase. This paper focuses only on what happens after a UCP-compatible checkout intent has been formed: the payment is authorized through an AP2-style mandate (or any equivalent user-authorization protocol that yields a verifiable, scoped mandate) and is then executed as a customer-to-customer wallet transfer. </div> <div> <br> </div> <div> We separate an agent that constructs or receives a UCP-compatible checkout intent, an authorization layer that converts a user-signed mandate into a wallet request, and a wallet system that is the sole execution authority for fund movement. We map each step to a concrete wallet API call, describe the trust boundaries and failure modes implied by the wallet state machine, and document the limitations of a closed-loop approach. </div> <div> <br> </div> <div> The core contribution is a single, narrow reinterpretation: the wallet customer-to-customer transfer is treated as the execution of authorized commercial intent - the settlement leg of a UCP checkout/payment intent - rather than as a peer payment. The model is UCP-compatible at the payment step and AP2-compatible at the authorization step; it depends on neither, and AP2 (or an equivalent scheme) can sit upstream as the source of mandates while the wallet-side execution discipline remains unchanged. This is a design/position paper: no implementation or empirical evaluation is presented. </div>