American Express launched a developer kit for AI agent verification, intent analysis, and payment credential checks. Visa had a six-month head start. Amex just closed the gap in a single announcement.

The protocol landscape just expanded again. And the competitive angle Amex is playing is different from anything we have seen.

When we published our definitive protocol map for Q1 2026 last week, American Express was already listed as a founding member of the x402 Foundation and a partner on Google's Universal Commerce Protocol. A participant, not a builder. That just changed. Amex published the Agentic Commerce Experiences (ACE) developer kit, and the scope is wider than anyone expected.

Agent registration and verification. Account enablement for consumers to register cards for agentic transactions. Intent analysis. Payment credential verification. Shopping cart context as a risk management signal. Consumer controls managed from the Amex app. And plans to bake rewards and benefits directly into agent transactions.

They have completed thousands of agentic payments in testing with AI platform partners. That is not a press release with a roadmap. That is a developer toolkit backed by live transaction data.

The ACE developer kit is designed to let partners integrate Amex payment capabilities into AI-driven shopping experiences. Here is what that means in practice.

American Express built six core capabilities into the kit. Agent registration creates a trust layer: before an AI agent can transact using Amex credentials, it gets verified. Account enablement lets cardholders explicitly register their Amex cards for agentic transactions, giving the consumer a deliberate opt-in rather than having an agent silently attach payment methods. Intent analysis evaluates what the agent is trying to do before the money moves.

Then comes the risk infrastructure. Payment credential verification confirms the card is valid and authorized. Shopping cart context feeds transaction details, what is being purchased, quantities, merchant category, into the risk engine as an additional fraud signal. Consumer controls let cardholders manage permissions directly from the Amex app.

Luke Gebb, executive vice president and global head of innovation at American Express, told American Banker: "We want to offer clarity for those who are asking AI agents to make purchases on their behalf."

That sentence understates the ambition. Amex is not just offering clarity. It is building the verification, risk, and rewards stack for agent-initiated transactions on its own rails.

Here is the part that separates Amex from open-loop networks.

Visa operates an open-loop network. It sits between issuing banks and acquiring banks. When a consumer's agent initiates a payment, the network routes the transaction, but the data is fragmented across issuers, acquirers, and processors. Building a complete picture of an agentic transaction requires coordination between multiple parties.

American Express runs a closed-loop model. It is the card issuer, the payment network, and the merchant acquirer. When an Amex cardholder's agent buys something from an Amex-accepting merchant, Amex sees both sides of the transaction. The consumer's authorization, the agent's intent, the merchant's inventory, and the settlement, all flow through a single infrastructure.

For agentic commerce specifically, that architecture creates a data advantage that open-loop networks cannot easily replicate. When the ACE kit analyzes shopping cart context as a risk signal, it can correlate that with the cardholder's spending history, the merchant's transaction patterns, and the agent's verification status. All in one system.

This is the competitive angle that nobody is talking about yet.

Visa has focused its agentic commerce infrastructure on trust, authentication, and fraud prevention. Those are the right problems to solve first. But Amex announced it will extend rewards and benefits to agent-initiated transactions.

Think about what that means for payment optimization. When an AI agent shops on your behalf, it will eventually choose how to pay. The simplest optimization is cost: pick the cheapest rail. Pay-by-bank through open banking. Stablecoins via x402. Debit over credit. That logic squeezes interchange revenue out of every transaction and routes volume away from premium card networks.

But total value is not the same as lowest cost. An agent optimizing for the consumer's best outcome needs to factor in rewards, cashback, purchase protections, extended warranties, and lounge access on travel bookings. American Express cards carry some of the richest reward structures in the industry. If those benefits flow through agent transactions, the payment optimization calculus flips.

An agent choosing the cheapest rail picks pay-by-bank. An agent optimizing for total value, price minus rewards, might pick Amex. That is a fundamentally different competitive dynamic, and it favors the premium network over the low-cost alternative.

Amex CEO Stephen Squeri framed it in the 2026 shareholder letter: "The winners in this new paradigm will go well beyond basic payment functionality by offering differentiated value, service and security."

He is betting that when agents choose payment methods, they will not optimize purely on price. He might be right.

We have tracked the agentic commerce standards race since last year. With Amex entering, the strategic contrast is sharpening.

Visa has the broadest infrastructure play. The Trusted Agent Protocol handles agent authentication, built with Cloudflare and Akamai for enforcement at the network edge. Intelligent Commerce Connect, launched this month, acts as a protocol-agnostic on-ramp supporting TAP, Stripe's Machine Payments Protocol, OpenAI's Agentic Commerce Protocol, and Google's UCP. And the Visa Acceptance Platform is already deploying through Fiserv across Europe. Visa's strategy is horizontal: connect to everything, process for everyone.

American Express is playing a different game entirely. Smaller acceptance footprint. Higher interchange. But the richest consumer base in payments and a closed-loop data advantage that open-loop networks cannot match. Amex is not trying to be the universal agentic commerce network. It is positioning as the premium one, where agents transact for consumers who value rewards, protections, and service over raw cost.

The approaches are complementary in some ways. An agent operating across merchant types might authenticate through Visa's TAP and still choose to pay with Amex for the rewards. The networks are not building mutually exclusive systems. They are competing for different layers of the same transaction.

This is not happening in a vacuum. Research from SoundHound and American Banker found that 70 percent of banking professionals believe agentic AI will have a significant or game-changing impact on the industry. But adoption is split: 64 percent of large banks are testing or deploying agentic AI, while only 38 percent of smaller institutions have started.

The top barrier? Security and compliance, cited by 51 percent of respondents. That is exactly the problem the ACE kit, Trusted Agent Protocol, and Agent Pay are designed to solve.

Over half of banking professionals said stronger regulatory guidance and seamless integration with existing tools would accelerate adoption. American Express, with its closed-loop model and direct issuer relationships, does not need to wait for third-party processor adoption the way open-loop networks do. It controls the full stack.

Three card networks now have production or near-production agentic commerce infrastructure. The protocol count in our landscape map goes from six to seven. The question is no longer whether card networks will support agent-initiated transactions. It is how agents will choose between them.

The rewards question is the one to watch. If Amex successfully extends Membership Rewards to agentic transactions, and if agents learn to optimize for total consumer value rather than lowest transaction cost, the competitive dynamics of agentic commerce shift toward premium networks. That challenges the assumption that machine-speed commerce will inevitably race to the cheapest rail.

EMVCo, the standards body jointly owned by the card networks, announced it is exploring how EMV specifications can support agentic payments. Amex is an active member. When the networks that compete for agentic commerce volume start writing shared specifications through EMVCo, interoperability moves from aspiration to requirement.

We started covering agentic commerce protocols when there were two. Then four. Then six, as we mapped in our definitive Q1 piece. Now seven. The stack is filling in faster than anyone predicted. The question that remains unanswered is the one that matters most: when an agent holds credentials from all three networks, what logic determines which one it uses?

When your AI agent holds cards from multiple networks, each offering different authentication, different protections, and different rewards, who writes the logic that decides which one wins?

Charlie Major is a Product Development Manager at Mastercard. The views and opinions expressed in Major Matters are his own and do not represent those of Mastercard.