- Mastercard has introduced Agent Pay for Machines and opened early access following pilot projects with Cloudflare, Coinbase and other partners.
- The system allows AI agents to pay for digital services within budgets and permissions set by companies.
- Mastercard said developers will receive early buyer and seller APIs, a toolkit, documentation and access to a sandbox environment.
Mastercard has introduced Agent Pay for Machines and showcased the payment system at the Cloudflare Connect conference, where it also announced an early access program for developers. The system allows AI agents to pay for digital services on behalf of companies without exceeding approved budgets and permissions.
Agent Pay for Machines uses the open x402 protocol, which allows digital services to request payment over the internet, along with Mastercard’s trust and security mechanisms. Agents can use the system to pay for data, AI models, computing resources and other digital services.
Mastercard said conventional payment arrangements typically require separate accounts, API keys, prepaid credits or billing relationships with each provider. Those requirements complicate transactions for agents that must independently select services while completing assigned tasks.
Digital vouchers authorize payments
Under the Agent Pay for Machines model, a company assigns a task to an agent and sets a spending limit. When the agent requests a paid service, the provider returns a 402 Payment Required status indicating that payment is needed.
The agent then creates a digitally signed payment voucher confirming a commitment to pay an amount within the approved limit. The provider verifies the voucher before delivering the service. These interactions occur off-chain, and providers can bundle the vouchers and redeem them later.
Mastercard said the model allows agents to purchase services without establishing a separate account with each provider, while companies retain control over authorized transactions and spending limits.
Mastercard demonstrates advertising use case
During a demonstration at Cloudflare Connect, marketing agency Hype instructed an AI agent to plan a test advertising campaign for the launch of a protein bar. The agent operated on an AI agent runtime platform that included Amazon Bedrock AgentCore.
The agent identified two service providers: AIsa for social media research and Cloudflare AI Gateway for image generation. After the budget was approved, it contacted the services. When Cloudflare Monetization Gateway requested payment, the agent signed a voucher, which the provider verified before continuing to process the request.
The agent received research from AIsa and eight advertising images from Cloudflare. According to Mastercard, the process did not require separate accounts, API keys or invoices for each provider.
Coinbase Developer Platform powered the wallet used to fund the agent, while Base provided the network for onchain settlement.
Early access opens to developers
Mastercard launched the Agent Pay for Machines Early Access program after pilot projects with Cloudflare, Ant International, AIsa, Base, BVNK, Coinbase and Hype.
The company said developers will receive early versions of buyer and seller APIs, a toolkit, documentation and access to a sandbox environment. Mastercard plans to develop the system further using feedback from customers and partners.
Mastercard describes Agent Pay for Machines as infrastructure connecting AI agents, payment providers, platforms and digital-service providers. The company aims to let agents independently pay for required resources while leaving spending controls with people and businesses.
Earlier, Citi and Bank of America restricted AI agents’ access to customer accounts.
Source: Incrypted
