Mastercard, Lobster.cash let AI agents pay with cards

On Thursday Mastercard and Crossmint's Lobster.cash announced consumers will soon be able to authorize AI agents to charge purchases to their existing payment cards. Transactions will be authenticated on Mastercard's network and governed by card issuers' authorization and fraud controls. The capability will begin with early access on OpenClaw, an open-source platform for autonomous agents.

The companies wrote that the system ties each transaction to a cryptographic record of a user's explicit approval so purchases can be traced to a specific authorization. Agents will not receive raw card credentials; they will use tokens and network-mediated controls that preserve existing cardholder protections.

Lobster.cash is an AI-agent payment layer built by Crossmint. The project connects crypto rails and payment partners including Solana, Circle, Visa, Mastercard, Basis Theory and Stytch. Crossmint raised $23.6 million last year in a funding round led by Ribbit Capital. OpenClaw is the developer-facing environment where agent-driven commerce applications can run.

Mastercard calls the integration Agent Pay. AI systems are moving beyond chat to automated buying, and payments firms have been developing standards and protocols; Visa introduced a Trusted Agent Protocol last year to set standards for AI-driven commerce.

Mastercard's chief digital officer Pablo Fourez wrote that integrating with Lobster.cash extends Mastercard's payments network and infrastructure and allows issuers to enforce limits and fraud protections while agents act on authenticated user directions.

The partners stated the initial rollout will test issuer controls and network-level authentication before the capability is expanded to more users and platforms.