Mastercard Debuts Agent Pay for Machines with Live Demo

Mastercard introduced early access to Agent Pay for Machines on October 20, 2026, following a demonstration that showcased an AI agent executing nine payments to two service providers on one approved budget. The initiative marks a step toward allowing autonomous software to handle routine digital transactions while maintaining human oversight on spending and permissions.

The system builds on the open x402 payment protocol, which supports HTTP 402 status codes for payment requests. In the demonstration at Cloudflare Connect, an AI agent from marketing agency Hype planned and completed a campaign for a protein bar launch. It purchased social media research from AIsa and generated eight campaign images via Cloudflare AI Gateway, all within limits set by the agency.

How Agent Pay for Machines Handles Authorization

Before any purchase, businesses define the AI agent's assigned task, spending limits, and permissions. When the agent encounters a paid service, the provider returns a 402 response. The agent then generates a digitally signed voucher confirming payment within its authorized range. Providers verify the voucher before fulfilling the request, enabling off-chain transactions that settle later in batches.

This approach separates authorization from settlement, allowing providers to group vouchers for redemption without each agent establishing individual billing relationships. Mastercard connects agent platforms, wallets, payment providers, and merchants through a shared trust layer to facilitate secure, high-frequency low-value exchanges.

Payment Infrastructure and Partners

End-to-end pilots involved multiple organizations. Coinbase supplied wallet infrastructure for funding the agent, while Base handled on-chain settlement. BVNK, acquired by Mastercard in August, acted as the agent wallet and managed payments and settlements in fiat or stablecoins through the Mastercard Global Money Platform. Cloudflare provided the monetization and AI gateways used in the demo.

Other participants in the pilots included Ant International, AIsa, and additional cloud and infrastructure providers. The setup supports purchases of data, computing power, APIs, and other digital services, with settlement occurring on-chain where needed.

Early Access for Developers

Mastercard released buyer- and seller-side APIs, along with documentation, a toolkit, and sandbox environments. Developers can now test agent-to-agent commerce scenarios and provide feedback to shape the roadmap. The platform remains in pilot stage and is not yet commercially available, with customer and partner requirements guiding future development.

Mastercard positions Agent Pay for Machines as part of a broader effort to create an interoperable trusted layer for AI-driven transactions, extending its recent work on stablecoin security and automated payment controls.

Broader Context for Autonomous Payments

As AI agents become more capable, they increasingly require the ability to discover services, agree to terms, and pay autonomously. Traditional systems often require separate accounts or keys for each provider, which complicates seamless operation. The new framework addresses this by focusing on authority and limits rather than direct control.

Pilots with more than 30 industry partners, including Stripe, Ripple, Polygon, and the Solana Foundation, have supported the technology since its introduction in June. The October 20 rollout at Cloudflare Connect highlighted real-world execution, with the agent completing purchases without human intervention after initial approval.