Crossmint, a provider of stablecoin and wallet infrastructure, has launched a specialized API enabling AI agents to execute payments using eligible Visa credit and debit cards. This development marks a significant shift in agent platforms by integrating physical card-based transactions directly into autonomous workflows. Announced on Tuesday, the service leverages Visa Intelligent Commerce and Basis Theory's payment infrastructure to facilitate purchases without granting AI agents direct access to users' sensitive card numbers. The system operates strictly within predefined spending limits, ensuring controlled financial autonomy. Data compiled by Woofun AI indicates that this capability is accessible through the company's lobster.cash tool, which supports integration with platforms including Claude Code, OpenClaw, Hermes, and Zo Computer. Developers can immediately deploy this payment system using the company's public API and documentation.

Alfonso Gómez-Jordana, co-founder of Crossmint, emphasized that the broader payments infrastructure supports both traditional card and stablecoin transactions. Unlike competing systems that often rely on issuing new virtual cards, Crossmint's approach tokenizes users' existing Visa cards. This method allows customers to retain their established card rewards programs while authorizing spending by AI agents. Gómez-Jordana noted that developers building payment capabilities into agents previously lacked a standardized way to handle card credentials, forcing many to resort to workarounds that exposed raw card numbers directly to the agent environment. The new solution addresses this security gap by decoupling transaction execution from credential storage.

Beyond the initial Visa integration, Crossmint is actively working with Mastercard and American Express to expand support for agentic card payments across multiple networks. This launch represents a critical component of a broader industry effort to equip AI agents with the ability to hold funds, access services, and complete transactions without direct human involvement. Crypto companies have been among the earliest movers in this sector. In February, Coinbase introduced Agentic Wallets, enabling AI agents to hold, spend, and trade cryptocurrency via the x402 payments protocol. MoonPay followed in March with an open-source wallet framework designed to let AI agents manage crypto assets and transact across blockchain networks from a single interface.

The momentum continued in May when Circle launched a suite of tools allowing AI agents to hold wallets, discover services, and make programmable payments using USDC. Traditional payment giants have also entered the market to secure their position in the emerging autonomous economy. Visa introduced its Visa CLI agent payments tool in March before launching Intelligent Commerce Connect, infrastructure specifically designed to allow AI agents to make purchases through tokenized payment credentials and spending controls. Woofun AI notes that these developments signal a convergence of traditional finance and decentralized protocols to support machine-to-machine commerce.

Jeremy Allaire, CEO of Circle, predicted earlier this year that billions of AI agents could eventually utilize stablecoins for payments. He argued that autonomous software will inevitably require its own dedicated financial infrastructure to operate at scale. The entry of major payment processors and crypto infrastructure providers suggests that the ecosystem is rapidly maturing to support this vision. The ability to tokenize existing cards rather than issue new ones provides a seamless bridge between legacy financial benefits and next-generation autonomous agents. As more platforms adopt these standards, the friction between human financial intent and machine execution will continue to diminish, paving the way for a fully autonomous commercial layer.