On July 8 Alipay released an Agent Infrastructure Suite and an Agent Marketing Suite for developers and merchants, plus tap to pay '碰一下' Agent nodes (tap terminals that hand the task to an AI agent) piloting on a Chinese campus and at a beer
On July 8 Alipay released a developer toolkit and a merchant suite that together let an AI agent find a service, request per-transaction user authorization, charge a card, and write a refund-able audit trail, with named pilots already live on a Chinese university campus and at a beer festival.
A LeiPhone launch report puts the offer in two acronyms. The Agent Infrastructure Suite (AIS) is the developer kit that turns the fragmented setup work for agent builders (account registration, API keys, service activation, top-ups, environment configuration, API debugging, scattered billing) into a supply network that the agent itself can discover, call, authorize, budget, and log. The Agent Marketing Suite (AMS) sits beside it as the merchant and retention layer. Alipay's own AI Pay product site and developer documentation are the canonical references for what is wired up live versus preview.
Two pilots already show the new loop working in the wild. Alipay's "Tap" feature, the 碰一下 tap-to-pay terminals already familiar in Chinese physical retail, now ships as Agent nodes, so a shopper can tap, hand the task to an AI agent, and let the bot finish. A Haier campus-laundry deployment sends a student's AI to wash a sweater on a recurring schedule. ITBear's coverage of the rollout describes a beer-festival pilot where one agent batched group orders and settled the tab. Sina Tech's framing treats both as live deployments, not staged demos.
On the transaction side, the suite's distinguishing piece is its authorization layer: per-transaction user authorization, budget controls, payment and refund handling for merchants, and transaction logs with risk and dispute traceability. When an AI agent spends real money, that combination answers who told the agent it could spend, how much, and on whose behalf, and what happens when the order goes wrong. The LeiPhone launch report lists all four, which is what turns the announcement from a developer convenience feature into a prototype of machine-spending trust.
Every named deployment is on mainland China, and AI付 is not marketed for international use. The source bundle provides no comparative data against Stripe, or Tencent, no agent adoption metrics, no merchant count, and no transaction volume, so claims about who wins agentic commerce are out of scope here. A 36Kr analysis describes the move as part of a Chinese payment-platform push toward agent readiness, alongside WeChat's parallel work in the same direction.
Alipay has put a recognizable consumer surface, the tap-to-pay terminal, behind a developer-facing protocol, with the Haier and festival pilots as the first two test cases. If recurring-schedule and group-order flows hold at campus and festival scale, the same 36Kr piece predicts similar agent-ready surfaces across Chinese retail by the end of the year.