Most of today’s payment systems, from Stripe to Alipay to crypto wallets, are fundamentally built for humans. They assume a user is present to approve, sign, verify, and complete a transaction.
But if we want autonomous agents to act on behalf of users, especially across the internet or onchain, things break quickly. Agents don’t “click pay,” they don’t pass CAPTCHAs, and they can’t navigate approval flows built for humans. Even crypto wallets today require explicit signatures, which assumes a conscious actor in the loop.
That’s why I’m wondering: should we be thinking of agent payments as a fundamentally different system? One with its own design principles, security assumptions, and UX flows?
Or are we still too early: is it just a subset of the existing payment stack, only needing better APIs and automation wrappers?
Curious how others think about this, especially builders working on agents, task execution, or payment infrastructure.