# Agent payments need a control plane

May 29, 2026

Being able to pay is only the beginning. Who gave permission, and what happens next?

Software should be able to buy services and settle obligations without a person clicking every button. But moving money is only one part of that job.

Payment standards solve useful problems. Builders still have to connect protocols, wallets, policies, and receipts. A successful payment shows that an agent can pay. It tells us little about whether the agent had permission.

The questions are simple: who authorized the action, what could the agent do, which rules applied, and what evidence remains?

A payment needs a control plane: bind the action to a clear mandate, check the rules before execution, ask for approval when needed, and leave evidence another person or system can verify.

The goal is a payment the user intended, within the authority they gave, with a result they can understand later.