The Google AP2 Payment Protocol Will Trigger Interesting Innovation

CANADA – 2025/07/25: In this photo illustration, the Google AI logo is seen displayed on a smartphone screen. (Photo Illustration by Thomas Fuller/SOPA Images/LightRocket via Getty Images)

SOPA Images/LightRocket via Getty Images

I think most people would now agree that the smart wallet, a digital wallet that can be controlled by AI agents, will be the central orchestration mechanism for next generation commerce. A new payments infrastructure is emerging to both re-energise past propositions (eg, micropayments) and create entirely new ones (eg, supply-chain currencies). This is why the announcement of the Google AP2 (Agent Payment Protocol) attracted so much attention from those of us developing strategies to exploit the trasnition to agentic commerce.

The Google AP2 Protocol

I think across the fintech sector we are looking forward to evolution of agentic commerce while being realistic about the practicalities. As many people have recognised, if we are going to have safe and secure agentic commerce then we must have a solid foundation of digital identity. Consumers, retailers and payments providers need to know that an agent is what it says it is and it has the authority to do whatever it is asking to do. The AP2 protocol is designed to provide interoperable guidelines as to how agents are authorised and held accountable for their actions. It is an open and non-proprietary protocol that builds on Anthropic’s Model Context Protocol (MCP) for connecting agents to external tools and Google’s Agent-2-Agent (A2A) protocol for, as you might imagine, inter-agent communications.

The new protocol uses open standards such as the World Wide Web Consortium (W3C) standard for verifiable credentials (VC) which means, as Simon Taylor rightly observed, that it is ready to obtain “internet scale” rather than being rooted in any single payment network or layer. This means new businesses springing up in a safe. secure and interoperable framework. That is, AIs will present their own verifiable credentials, with our consent, to get things done. As I commented here before, my agent should not show up at a bank pretending to be me: it should show up at a bank as a agent acting on my behalf.

(This means, of course, that we need a digital identity framework in which agents have their own identities, meaning we are entering an era where know-your-agent, or KYA, will be central to fintech evolution.)

AP2 has some broad support, having been developed in partnership with companies ranging from Adyen and to PayPal and Coinbase. The goal is to provide a standardised way to allow agents to shop and pay. At the heart of the transactions are what Google calls “mandates”. In essence, agents need to have two madates to effect a transaction: an “intent mandate” (I am authorised to look for return flight to Boston for David G.W. Birch) and a “cart mandate” (Yes, go ahead and book that flight, here’s an Amex token to charge it to).

Ap2 is a big deal and it seems to me that the protocol and its ecosystem will reshape the very nature of e-commerce, but I am not an expert on retail strategies, so I called someone who is. Online retail expert Eva Pascoe, who co-founded the first Internet cafe in the UK and later brought major fashion brands to the web and mobile shopping, told me that while this move will undoubtedly benefit consumers, it means that retailers will need to both rethink their online strategies (or risk being reduced to low added value fulfillment) and to accelerate their thinking and their “skunkworks” exploring agent opportunities. With Shopify and Etsy already in the ecosystem, I am sure that she is right to suggest that retailers need to see agentic commerce within strategic horizons and not a science fiction dream.

As an aside, in many ways AP2 reminds me of the Internet Engineering Task Force (IETF) “Internet Open Trading Protocol” (IOTP) standard that I contributed to more than 20 years ago. In that case, as I am sure will be true in this case, sorting out the message formats and digtial signatures was the easy part. As Tom Noyes points out in his (excellent) summary of AP2 and its issues, the real challenge is that identity and payment must all operate within well-defined agreements (ie, there must be governance) to manage financial and regulatory risk. This is spot on, and it is why the presence of players such as Mastercard and American Express (who have many many, many years experience of such things) means that serious players are looking at AP2 as a place to begin their agentic commerce experiements.

More Consequences of Google AP2

One particular aspect of the launch caught my eye. AP2 is payments agnostic. It is not built around payment cards and has been designed to be extensible. While the focus is currently on “pull” payments such as credit cards it will support “push” payments such as UPI in India and PIX in Brazil). Thus the protocol is future-proofed and in theory new payment products (a central bank digtal currrency, to pick an obvious example) can just plug in without a redesign.

However, account-to-account and Dogecoins to one side, I thought that the fact that Google has already developed an extenstion (working with Coinbase, MetaMask and the Ethereum foundation) to integrate the x.402 protocol, which means that agents can already transact using stablecoins, is rather interesting and likely to bring the smart wallet to the fore. This may well turn out to be the most interesting development in the payment space for some time.

The reason for this is obvious. Agents cannot have bank accounts, but they can have smart wallets that they use to store stablecoins and to spend them via AP2. That means an environment for permissionless innovation and it seems to me that the state of the art will progress in this direction, with permissionless experimentation and “natural selection” showing us how the new economy can develop.

The Google AP2 announcement is another step towards an economy dominated by machine actors, with agents using smart wallets to pay other agents directly for the services they require. Stablecoins satisfy these new payment needs and will stimulate the evolution of an infrastructure for the transfer of digital assets. Now, that is not to say that other payment solutions will evolve to support robot-to-robot (R2R) payments in time