Cleverbridge Completes France's First Live AI Agent Payment with Visa and Revolut: How Trusted Agent Protocol and Payment Passkey Meet Europe's SCA Rules

Who Did What in France's First Agentic Payment

Cleverbridge Completes First Live AI Agent-Initiated Payment in France with Visa and Revolut

Cleverbridge completed the first live AI agent-initiated payment in France, with Visa and Revolut, on a Passkey-authenticated consumer card.

www.businesswire.com

On September 24, 2026, Cleverbridge, a managed global commerce platform for software and technology companies, announced that it had completed France's first Passkey-authenticated agentic payment in a live checkout. The transaction was part of a pilot with Visa and Revolut announced earlier in the month.

Visa's test agent, My Agent, initiated the purchase. Cleverbridge recognized it as an approved agent through Visa's Trusted Agent Protocol and completed the checkout. A Visa Payment Passkey authenticated the transaction, and Revolut authorized the payment on a French consumer card. The whole flow ran on Visa Intelligent Commerce (Visa's umbrella of AI commerce initiatives) using Visa's existing payment infrastructure.

Founded in 2005, Cleverbridge processes more than $1 billion a year across 240+ markets, with clients including Autodesk, SUSE, Tenable, Veeam and Shure. It takes on payments, billing, tax and compliance on behalf of software companies as a reseller, and in this transaction Cleverbridge itself was the "merchant."

Condition One: The Merchant Knows Which Agent It Is Dealing With

At the center of the announcement is this statement from CEO Richard Stevenson.

"

Here the merchant knew which agent it was dealing with and the customer's bank stayed in charge of the payment. That is what has to be true before agentic commerce can carry real volume.

Source: Richard Stevenson, CEO, Cleverbridge

The first condition rests on the Trusted Agent Protocol (TAP), which Visa released with Cloudflare in October 2025. When an AI agent visits a merchant's site, it attaches an agent-specific cryptographic signature indicating that it is a trusted agent with an intent to buy. The merchant checks this against a Visa-maintained list of approved agents (Cleverbridge calls it the Agentic Directory, Visa the Agent Directory) and can separate legitimate agents from malicious automated traffic.

Why is this a prerequisite? Traditional e-commerce sites are built to block non-human automated traffic as bots. Once agents start shopping, that defense also stops legitimate purchases. Letting everything through, on the other hand, opens the door to fraudulent automated buying and credential theft. Without a way to turn "unknown automated traffic" into "a verified counterparty", merchants cannot decide to accept agent orders at all.

News from the same day shows this is not theoretical. According to The Paypers, Amazon blocked Meta's AI agent Muse from shopping on Amazon.com. Amazon said Meta had not notified it and that Muse does not identify itself when browsing. Meta has said Muse has no visibility into users' passwords or payment methods.

Side by side, the contrast is sharp. An agent that would not identify itself was shut out; an agent that proved its identity completed a transaction. In its July announcement, Cleverbridge said TAP lets it define how approved agents can access its platform and product information within its existing risk frameworks. The real value for merchants is not a blanket yes or no, but the ability to set access rules per counterparty.

Condition Two: The Issuing Bank Stays in Charge

The second condition is tied closely to European regulation. The EU's Payment Services Directive (PSD2) requires Strong Customer Authentication for online payments. SCA verifies identity with at least two of three factors: something you know, such as a password; something you have, such as a phone; and something you are, such as a fingerprint or face. The rule assumes a human acting in the moment, and applying it when an agent buys on someone's behalf has been an open problem.

Visa's answer is the Visa Payment Passkey, which lets cardholders approve payments with biometrics much as they unlock a phone, with the biometric data never leaving the device. Crowdfund Insider reports that in this transaction the Passkey bound the payment to a verified cardholder and an explicit prior instruction. No one typed card numbers or pressed an approval button at checkout. Tokenization replaced the card number, so the agent never handled raw card data.

This is where "the issuer stays in charge" comes in. Revolut retained authorization over the transaction as with any card payment, and Visa's standard protections such as fraud monitoring remained in place. The Paypers describes the pilot's purpose as establishing whether an agent-led debit could work within existing SCA, tokenization and issuer-control frameworks, rather than introducing a separate payment rail.

Cleverbridge's release includes one more pointed claim. Visa has said that passkeys guarantee purchase intent, protecting consumers from agent mistakes during the purchase and protecting merchants from AI-led chargeback requests. A chargeback reverses a payment after a cardholder dispute, and the merchant loses the sale. How to handle claims that "the AI bought it without asking" is a major merchant concern, and this claim is aimed squarely at it. It is, however, Visa's position as relayed by a merchant, not a published dispute rule.

How Far Does "France's First" Reach?

The "first" in the headline comes with qualifiers. The Paypers notes that this is the first instance in France involving Visa and Passkey authentication, not the first agent-initiated payment in Europe. That earlier milestone came in March 2026, when Santander and Mastercard completed a transaction on Mastercard Agent Pay, a separate scheme stack.

Within Visa's own track, this transaction is not an isolated event either.

Visa launched the Agentic Ready programme in Europe in March, and at the Visa Payments Forum on July 2 it announced that more than 30 European issuers had completed agent-executed transactions. Revolut was already on that issuer list, and Cleverbridge on the merchant list. Biometric Update reported the same month that Worldline, ING and Visa had completed an agent-driven payment authenticated with a Payment Passkey.

What is new here is less a national first than a step up in maturity. In July, Cleverbridge said it had completed agent-initiated transactions "in its own environment." This time it took that work into the market with a live consumer card. When Stevenson says it "has to work in a live market rather than a lab," this is the difference he means.

It is still not something consumers can use. Both The Paypers and Crowdfund Insider report that the functionality has not been activated in the Revolut app and that the transaction was part of a controlled pilot. The purchase was initiated by Visa's test agent, not a general-purpose assistant such as ChatGPT or Gemini.

Why a Software Reseller Moved First

It is telling that the first merchant was a software reseller. Cleverbridge says support for approved AI agents is being built into the commerce environment it operates rather than becoming a separate project each client has to run. Software companies selling through Cleverbridge move closer to accepting agent orders without building anything themselves.

For many merchants, implementing TAP verification and agent access policies one by one is a heavy lift. Visa's July release likewise said infrastructure providers such as Cloudflare and Akamai support implementation. The pattern emerging is that agent readiness spreads first through the layers that aggregate many merchants: payment service providers, resellers and CDNs, rather than individual stores.

Questions Still Open

A successful pilot does not settle every operational question. The Paypers lists the unresolved issues: the scope of an agent's mandate, merchant allow-list management, spending caps, chargeback liability when the actor using the card is software rather than a person, and how Passkey binding applies to an agent session as opposed to a browser session.

Items not disclosed in this announcement:

• Product purchased, amount, and number of transactions: not disclosed
• Specific rules for allocating chargeback liability: not disclosed
• Timing of general availability in the Revolut app: not disclosed
• Timing of support for purchases from general-purpose AI assistants: not disclosed
• Merchant-side implementation costs or fees: not disclosed

What Merchants Should Take From This

The transaction shows where the design of agentic payments is heading.

First, the shift is from "blocking" agents to "identifying and admitting" them. Leaving bot defenses as they are risks losing orders from legitimate agents too. A decision like Amazon's to refuse agents that will not identify themselves also only becomes a principled line once identification mechanisms like TAP exist. It is worth checking whether your bot management and WAF settings can distinguish signed agents.

Second, authentication and authorization stay with the card networks and issuers. Japan, for example, requires EMV 3-D Secure authentication for online card payments, which, like Europe's SCA, assumes the cardholder is present and acting. The Passkey approach Visa has shown in Europe is a useful reference for markets with similar authentication rules.

Third, there is a build-or-rely decision. When commerce platforms and payment providers build agent support in, as Cleverbridge has, the burden on individual merchants shrinks. A realistic first step is to ask your cart and payment providers how they plan to support TAP and agentic payments.

Conclusion

The Cleverbridge, Visa and Revolut transaction shows agentic payments moving from "does it work technically" to "does it work within regulatory and liability frameworks." The two conditions, that the merchant knows the agent and the issuer controls authorization, have been shown to coexist under Europe's strict authentication rules.

The next question is whether the same transaction can happen from a general-purpose AI assistant rather than a test agent, and from an ordinary app rather than a pilot. Just as important is how clearly the rules will spell out who bears the cost when a cardholder says "the AI bought it without me."