The Agent Card: Mastercard’s Bet That Your AI Will Buy Better Than You Do

(SeaPRwire) – By: Ethan Gallagher
The core issue is not whether the technology works. It does. The real problem is who pays when the AI screws up. Mastercard and Alchemy are launching a system where an AI agent holds a virtual card and spends on your behalf. No more confirming every purchase. You set the rules, the bot does the clicking. It sounds convenient. For now, it is just a risky experiment.
Look at the official release. Mastercard partnered with San Francisco-based Alchemy to link its agentic payment framework with Alchemy’s AgentCard infrastructure. The tool launched in March 2026. Visa joined the party in June. Users connect their AI agent to their Mastercard account via Alchemy. They set spending caps. They restrict retailers. They can even require a check-in before a transaction. The tech relies on “agentic tokens.” These are generated by the issuing bank. They bundle the cardholder’s intent and transaction details. The network verifies the agent stays within limits. Mastercard’s Chief Product Officer Jorn Lambert said it is not about if, but when.
Now, strip away the PR. What is actually happening here? The major card networks are not building the agents. They are building the pipes. Alchemy is a private startup. It is not publicly traded. Mastercard and Visa stocks were flat on the news. Why? Because this is a feature, not a product. The real value capture lies in the issuing banks. They generate the tokens. They bear the fraud risk. The card networks get to maintain their position as the universal gateway. If you let an AI buy anything, the barrier to entry for fintechs drops. But if you require a bank-issued token for every agent, you keep the money flowing through your rails. The “agentic token” is the lock. It ensures that no matter which AI app you use, the payment still hits a Mastercard or Visa line.
The supply chain reality is blunt. Trust is the bottleneck, not the code. Brendan Coughlin of Citizens Financial Group called it a risk. Regulatory questions remain unresolved. Who is liable if the agent hallucinates a purchase? Who pays if the agent acts outside its boundaries? Until those answers exist, this is a playground for the wealthy and the tech-savvy. For the average consumer, this is a way to lose control of their wallet. The landscape is shifting from “human clicks” to “algorithmic intent.” The banks are holding the levers. The agents are just the hands. Until the liability models are clear, the mass market will stay on the sidelines. The infrastructure is ready. The trust is not.