Agentic AI Gets Its Own Payment Protocol
A traveler books a flight through an artificial intelligence (AI) assistant embedded in a super app. The agent finds the fare, selects the seat and completes payment. The user never leaves the interface.
The digital wallet was built for a human tapping a screen. It was not built for software acting on their behalf.
At its MoMents 2026 FinTech forum in Kuala Lumpur, Ant International introduced the Agentic Mobile Protocol, or AMP, the world’s first agentic payment framework built for mobile interfaces including digital wallets, super apps and wearable devices. The protocol is open-sourced.
Existing AI payment infrastructure largely runs on card rails not engineered for agents initiating, authenticating and settling transactions without a human in the loop. AMP, however, was.
A Protocol Built for Where Commerce Already Lives
Digital wallet users reached 4.4 billion in 2025. That number is expected to exceed 6 billion by 2030. The payment infrastructure serving those users was not designed for AI agents. AMP is built to close that gap.
Merchants, AI platforms and agent builders can embed payment capabilities directly into existing workflows. No system overhaul is required. AMP cuts the number of steps required to link a payment agent to a digital wallet by 50% compared to traditional card-binding methods, according to Ant’s press release.
The network behind the launch is substantial. Alipay+ connects more than 40 digital wallet partners. It covers 1.8 billion user accounts and 150 million merchants globally, the company said. Ant is among the first partners of Mastercard and Visa to pilot card-based transaction capabilities for AI agents. The company is also working with Google on broader agentic commerce protocols.
Identity Is the Infrastructure
The buyer in an agentic transaction is no longer the person holding the phone. Banks, processors and networks need a way to verify that a software agent is acting within the scope of what a user authorized. AMP addresses that directly.
A “Know Your Agent” framework establishes a digital identity for each AI agent. It certifies what the agent is authorized to do. A separate Agent Trust Rating system functions as a dynamic risk tool. It determines whether an agent is trustworthy and controls how much autonomy it is granted. Every agent-initiated transaction carries a money-back guarantee for payment partners in the event of an account takeover.
Without that identity layer, agentic commerce does not scale. It becomes a fraud surface.
What the Payments Stack Needs to Do Next
Ant’s announcement arrives as Stripe separately embeds checkout into AI-native surfaces. Stripe is enabling stablecoin payouts to creators on Meta. It is also connecting to Google’s AI Mode through its Universal Commerce Protocol.
Stripe is building from the discovery and merchant side. Ant is building from the wallet and mobile-first side. The two efforts are not coordinated. Whether their standards interoperate is an open question the industry has not answered.
The global agentic commerce market is projected to reach around $28 billion by 2030 at a 46% compound annual growth rate, Ant said in its release. The volume is coming. The identity, consent and settlement standards that will govern it are still being written by companies working independently. For banks, processors and payment networks, that is the infrastructure problem in front of them now.
Merchants will need product catalogs that agents can read and act on. Processors will need new authentication flows that verify software, not just cardholders. Networks will need dispute and fraud frameworks built for transactions where no human tapped a screen. The companies that solve those three problems first will not just participate in agentic commerce. They will set its terms.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
The post Agentic AI Gets Its Own Payment Protocol appeared first on PYMNTS.com.