AgentCard Gives AI Agents Their Own Identity. That Is the Point.

Alchemy’s integration of AgentCard with Mastercard Agent Pay establishes a dedicated identity layer for autonomous software. By provisioning each agent with a unique email address, phone number, stablecoin wallet, and tokenized payment credentials, Alchemy is attempting to treat AI agents as distinct, verifiable entities rather than mere extensions of a user’s existing account. This is not a payment product dressed up as identity — it is the reverse.

When AgentCard launched in June 2026 with Visa Intelligent Commerce, it established the baseline for this architecture. The current Mastercard integration adds a critical component: Verifiable Intent. Co-developed with Google, this standard provides a tamper-resistant cryptographic proof that links a consumer’s specific instructions to a transaction outcome. It allows for selective disclosure, ensuring that only the minimum required information is shared during the checkout process.

The technical implementation is designed for speed — a single command-line instruction and under one minute to set up. Once active, the system encrypts credentials at rest and provides granular spending controls, including merchant category restrictions, geographic limitations, and instant freeze-or-revoke capability. These features address the operational friction our beat has tracked: the current reality where only 3% of transactions involve agents and developers face integration costs ranging from $5,000 to $500,000 per protocol, as our analysis of the protocol proliferation tax documented.

Alchemy CEO Nikil Viswanathan framed the shift plainly: “AI agents are quickly moving from simple assistants to software that can take action for people. To make that possible in everyday commerce, agents need a trusted way to pay, prove authorization, and operate within clear user-defined limits.” Sherri Haymond, Mastercard’s EVP of Digital Commercialization, pushed the framing further: “The future of commerce isn’t just about agents that can act — it’s about agents that can be trusted to act on your behalf.”

The infrastructure is being built, but the environment remains fragmented. Visa completed “hundreds” of secure agent transactions in a closed beta and projects “millions” by holiday 2026. Between those numbers sits a trust deficit that identity provisioning alone does not close: only 23% of consumers trust AI to handle payment transactions, and 85% demand transparency regarding how their data is used, according to Visa’s Earning Trust report.

The corporate environment presents a parallel disconnect. While 90% of executives express confidence in their visibility over AI usage, only 11% of applications are actually visible to IT departments. Because 67% of US workers are already using unapproved AI tools, the introduction of a formal identity layer does not automatically bring these unauthorized agents into a managed, secure framework. AgentCard provisions identity for agents that opt in. The shadow agent problem is about the ones that do not.

Liability remains another unresolved friction point. Merchants are currently hesitant to engage with agentic commerce — 93% believe the AI provider should bear the financial risk of transactions, and only 28% are willing to offer their full product range to agents. The technology provides the mechanism for the transaction. It does not settle the question of who is responsible when an agentic transaction fails or results in a dispute.

Alchemy’s work with Mastercard and Visa provides the plumbing for identity and authorization — addressing the technical requirements for agents to operate in a commercial environment. But the structural issues — consumer trust, shadow agent proliferation, and merchant liability allocation — are not solved by the provision of a dedicated email address or a tokenized credential. The industry is moving from experimentation to a more formal infrastructure. The transition to everyday use remains contingent on factors that code alone cannot resolve.