Visa, Mastercard, Stripe, and Adyen are building the rails for agentic payments. What this concretely changes for e-commerce merchants in 2026.
An AI agent buying on your behalf doesn't carry a credit card in a physical wallet. It also has no persistent browser session to re-enter a 3D Secure code. That's precisely the problem Visa, Mastercard, Stripe, and Adyen are trying to solve right now: how to authorize, secure, and trace a payment when the buyer is no longer a human sitting in front of a screen.
Agentic payments refer to the set of protocols that enable an AI agent to initiate and validate a commercial transaction on behalf of a user, without manual intervention at each step. This isn't science fiction: according to Forbes, major payment networks have been actively building this infrastructure since early 2026, in a race to become the de facto standard.
We've already explored preparing Shopify stores for buyer agents in Agentic Commerce on Shopify, on the catalog and structured data side. Here, we go down a level: the payment layer itself, where things get really uncomfortable for merchants.
The problem classical checkout can't solve
An e-commerce payment funnel is designed for a human. It assumes a browser with cookies, a moment to think between adding to cart and confirming payment, and above all the ability to recognize a CAPTCHA. An AI agent that negotiates and buys in seconds breaks all three assumptions at once.
The real issue isn't technical. Webhooks and REST APIs have known how to communicate between systems for a long time. The real issue is trust. How does a merchant know that the agent placing an order is acting with the genuine consent of the user, and not beyond the budget or preferences set for it? No current payment protocol has a standardized answer to this question.
Who's building what: four active projects
According to Forbes, major payment players—Adyen, Stripe, Visa, and Mastercard—are locked in a race to build the necessary infrastructure for agentic commerce. Adyen notably launched an offering called "Agentic," designed as a universal transaction layer capable of serving multiple agent types without dependence on a closed ecosystem.
| Player | Approach | What It Means for a Merchant |
|---|---|---|
| Adyen | Universal transaction layer ("Agentic") | Multi-agent interoperability, less lock-in |
| Stripe | Extension of existing API infrastructure | Probably faster integration for merchants already on Stripe |
| Visa | Dedicated network rails for AI-initiated transactions | Requires network adoption, scale effects |
| Mastercard | Network approach similar to Visa, direct competition | Standards battle, fragmentation risk |
This fragmentation is no small detail. If Visa and Mastercard each impose their own "proof of mandate" format for agents, merchants will need to integrate multiple protocols in parallel—much like the period when every mobile wallet had its own SDK before standards stabilized.
What this changes for a shopify merchant today
Concretely, if you sell on Shopify, two worlds will need to talk to each other: your existing checkout stack, and this new agentic authorization layer. Shopify Functions already lets you customize checkout logic on the merchant side. But verifying that a transaction comes from a properly mandated agent, not a malicious bot impersonating that mandate, isn't covered by any standard Shopify API to date.
A concrete example to illustrate the stakes: an e-commerce site selling recurring subscriptions will need to decide whether an AI agent can auto-renew a subscription beyond a certain amount, or whether it must go back to human confirmation. That threshold doesn't exist in any protocol today—every merchant will have to define it themselves, absent an imposed standard.
And that's where it gets interesting for technical teams. Validation logic can no longer live solely on the front end. It must live on the server side, in a layer that can distinguish a human request from an agentic request—much like the agent gateways we discussed for AI agent production in general, but applied specifically to payments.
The risk no one has solved yet: responsibility
Jenn Tejada, executive chair of PagerDuty, told Forbes in early July 2026 that AI is moving from experimentation to production, and agentic systems introduce specific failure modes, including model drift—a progressive shift in a model's behavior from its original intent.
Applied to payments, this risk takes on a very concrete form. An agent that drifts slightly in interpreting a budget can multiply micro-purchases that individually stay below the radar of classical fraud detection, but that cumulatively represent real losses. None of the protocols announced by Visa, Mastercard, Stripe, or Adyen has communicated in detail about who—the platform, the merchant, or the card issuer—bears financial responsibility in case of such drift.
That's an honest limitation to state: building the technical infrastructure for agentic payments is simpler than settling the legal and contractual question of responsibility. The two are advancing at very different speeds, and the latter risks slowing adoption far more than the former.
How to prepare without breaking everything
No need to rewrite your payment stack tomorrow. But a few reflexes are worth adopting now. First, audit your payment webhooks to identify which ones might one day receive requests signed by an agent rather than a human browser. Next, define internal thresholds beyond which human validation remains mandatory, independent of what protocols will later impose. Finally, follow announcements from Adyen, Stripe, Visa, and Mastercard on their agentic APIs—whichever standard wins this battle will determine the least costly integration in the medium term.
The bottom line
Agentic payments are no longer a lab hypothesis: Adyen, Stripe, Visa, and Mastercard are actively investing in it as of 2026. The standards battle between these players will probably fragment the market before stabilizing, as has happened with other payment technologies before it. And the real bottleneck isn't technical—it's the question, still unresolved, of responsibility when an agent drifts.
If your Shopify store starts receiving traffic from AI agents, it's better to anticipate server-side validation logic now rather than discover it in production. Contact fstck.co to assess your exposure to these new payment flows.
Frequently asked questions
What exactly is agentic payment?
Agentic payment refers to protocols that enable an AI agent to initiate and validate a financial transaction on behalf of a human user, without the user manually confirming each payment step. It relies on a form of digital mandate issued by the user to the agent.
Does shopify already offer a native solution for agentic payments?
No, not as of now. Shopify Functions allows you to customize checkout logic, but no standard API yet exists to natively distinguish between a human payment request and one initiated by a mandated AI agent.
How can a merchant protect themselves against an ai agent that drifts beyond its original budget?
By enforcing mandatory human validation thresholds beyond a certain amount on the server side, independent of the payment protocols used. This is the only reliable protection while payment networks clarify responsibility rules in case of model drift.
What's the difference between adyen's approach and visa or mastercard's on agentic payments?
Adyen built a transaction layer designed as universal and interoperable with multiple agent types, while Visa and Mastercard are each developing their own network rails, risking fragmentation of standards in the short term.
Does agentic payment concern only large e-commerce merchants?
No. Any merchant exposing an API or catalog accessible by an AI agent—even a small Shopify store—can receive agentic payment attempts. Company size doesn't protect against this exposure; only the validation logic you put in place does.
