Mastercard Gives AI Agents Virtual Cards for Autonomous Shopping

AI shopping assistants are moving beyond recommendations and into actual transactions. Mastercard has partnered with Alchemy to integrate Mastercard Agent Pay with Alchemy’s AgentCard, allowing AI agents to receive virtual payment credentials and make online purchases on behalf of users.

The new system is designed to let AI agents handle tasks such as ordering products, purchasing services, booking travel and managing subscriptions without requiring users to manually approve every individual transaction. Users can establish spending limits and other controls before allowing an agent to act.

The development marks another step toward agentic commerce, where AI systems can independently discover products, make decisions within predefined instructions and complete purchases.

How Mastercard AgentCard Works

Alchemy’s AgentCard provides AI agents with a dedicated payment identity that can be used for online transactions. The service can issue virtual cards with spending limits, allowing users to control how much an agent can spend.

According to AgentCard, users can create a single-use virtual card for a specific amount. The company’s current setup process allows developers to provision an agent’s payment infrastructure through a command-line interface.

The Mastercard integration adds Mastercard payment credentials to this infrastructure.

Instead of giving an AI agent unrestricted access to a user’s everyday payment card, the system is designed around controlled credentials. Mastercard says its broader Agent Pay framework supports credentialing, permissioning, transaction processing and settlement, allowing organizations to establish rules around automated payments.

AI Agents Can Make Purchases Without Constant Approval

Traditional online shopping requires a person to select a product, enter payment information and confirm the transaction.

Agentic commerce changes that workflow.

A user could instruct an AI agent to find a suitable product within a specific budget, for example. The agent could then search available options, compare prices and complete the purchase using an authorized payment credential.

Mastercard and Alchemy’s system is designed to support this type of autonomous purchasing while maintaining user-defined controls. Spending limits and merchant restrictions can determine what an AI agent is allowed to purchase.

This means the agent does not necessarily need to ask the user for permission every time it reaches checkout.

Mastercard Is Building Infrastructure for Agentic Commerce

The Alchemy partnership is part of a broader Mastercard strategy around AI-powered payments.

In June 2026, Mastercard introduced Agent Pay for Machines, a service designed for programmatic transactions between machines and AI agents. The company said these systems could eventually support extremely high volumes of automated payments, including transactions worth very small amounts.

Mastercard’s Agent Pay architecture includes several important components:

• Credentialing: Agents receive identifiable payment credentials.
• Permissioning: Spending rules and authorization limits can be enforced.
• Transacting: Verified agents can conduct payments across supported systems.
• Settlement: Transactions can be settled through different payment rails.

The company says its approach is intended to bring existing payment-network controls into an environment where software, rather than humans, increasingly initiates transactions.

Verifiable Intent Could Become Critical

One of the biggest challenges with AI-powered shopping is determining whether a transaction genuinely represents the user’s wishes.

Mastercard has developed Verifiable Intent as part of its agentic commerce strategy. The technology is designed to connect an agent’s transaction with authenticated user authorization and the relevant purchase information.

That could become important if an AI agent makes a mistake.

For example, if a shopping agent purchases the wrong product, exceeds its intended budget or interacts with an unauthorized merchant, payment networks need mechanisms to determine what the user actually authorized.

Mastercard says its agentic payment infrastructure is being built around trust, traceability and user controls.

Visa, Mastercard and Ant Are Working on “Know Your Agent”

The move toward AI-powered payments is not limited to Mastercard.

On September 10, 2026, Ant International, Mastercard and Visa announced work on a “Know Your Agent” interoperability framework intended to help identify and authenticate AI agents across payment networks, wallets and marketplaces.

The framework is intended to provide greater visibility into who operates an AI agent and which human or organization is responsible for it.

The concept is similar in spirit to “Know Your Customer” processes used in financial services, but adapted for autonomous software.

The proposed approach includes agent identification, certification requirements, traceability and ongoing monitoring. These mechanisms could allow financial networks to revoke or restrict an agent if its behavior changes or violates predefined requirements.

Security Remains a Major Challenge

Giving an AI agent access to money creates obvious security questions.

An AI system may be able to make decisions quickly, but it can also misunderstand instructions, encounter malicious websites or be manipulated by external content.

A compromised agent could potentially attempt transactions outside its intended purpose if payment credentials and permissions are not properly isolated.

That makes spending limits, merchant restrictions, authentication and rapid card revocation important parts of the infrastructure.

AgentCard says its platform includes network-enforced spending limits, real-time transaction visibility and the ability to freeze or revoke cards.

The larger industry challenge will be ensuring that these protections continue to work as AI agents become more autonomous and capable.

What Agentic Commerce Could Look Like

The Mastercard-Alchemy partnership points toward a future where the checkout process becomes less visible to consumers.

Instead of opening multiple websites and completing separate payment forms, users could simply give an AI agent a task such as:

“Find the best option under my budget and purchase it.”

The agent could handle discovery, comparison and payment while operating within predefined rules.

Mastercard is also developing merchant-facing tools designed to help businesses make products discoverable to AI agents and support agent-powered shopping experiences.

That could eventually create an ecosystem in which AI agents act as intermediaries between consumers and merchants.

The Future of AI Payments

Mastercard’s partnership with Alchemy demonstrates that AI agents are increasingly being treated as participants in the digital economy rather than simply software assistants.

The important development is not that AI systems can technically enter payment information. It is that financial networks are building identity, authorization and payment infrastructure specifically for autonomous agents.

For consumers, this could make online shopping significantly more automated. For businesses, it could create new ways to sell products and services through AI platforms.

But the success of agentic commerce will depend heavily on trust.

As AI agents gain the ability to spend money independently, payment networks will need to prove that every transaction can be traced to an authorized agent, a defined set of permissions and ultimately a real person or organization.

The Mastercard-AgentCard integration is another major step toward that model—and it suggests that the future of online shopping may involve far fewer checkout buttons and far more AI agents making decisions on our behalf.