Inside Agent Pay for Machines: Mastercard's Bid to Power the AI Economy

Mastercard has formally entered the race to become the financial backbone of the AI economy. On June 10, 2026, the company announced Agent Pay for Machines (AP4M), a new service designed to let AI agents transact with each other at machine speed, without a human hovering over an approve button. It’s a deliberate pivot from the company’s 2025 Agent Pay program, which focused on human-initiated payments, to a future where software bots handle the checkout flow themselves. The launch positions Mastercard against a crowded field of competitors—from Visa and Google to crypto-native protocols like Coinbase's Base network—all vying to control the payment rails for agentic commerce.

What is Agent Pay for Machines?

AP4M is a service for automated, high-frequency, low-value transactions executed by AI agents across Mastercard’s global network. The framework is purpose-built for a world where an AI assistant might need to pay a fraction of a cent to access a data feed or settle a logistics fee without any manual approval.

The new service provides a structured way for network participants to apply traditional payments trust and governance controls to this emerging machine-driven commerce. The announcement describes transactions that are "permissioned, orchestrated, and settled at machine speed."

The Four-Step Framework for Autonomous Payments

To make autonomous agent payments trustworthy, AP4M rests on four core pillars that move a transaction from a human's instruction to final settlement.

• Credentialing: Every AI agent is given a verifiable identity through a mechanism called Verifiable Intent. This isn't a simple login; it’s a cryptographic proof that allows any participating party to recognize and trust the agent across different ecosystems without sharing a private database.
• Permissioning: Before an agent can spend a single cent, the controlling organization can programmatically set and enforce authorization rules, such as specific spending limits, approved counterparties, or transaction velocity caps.
• Transacting: Once credentialed and permissioned, verified agents can discover each other and transact programmatically across different service providers and systems.
• Settling: The actual movement of value happens across Mastercard’s multi-rail network, giving organizations a choice of payment rails for each transaction.

Verifiable Intent: Moving Agent Authorization to the Blockchain

The Verifiable Intent framework is the cryptographic core that makes AP4M more than just another software integration. Instead of logging a user's spending authorization in a single, private database that only one company can see, Mastercard stores these permissions on public blockchains.

This decentralized approach means an AI agent’s mandate to act—"buy a domain name and set up hosting for my flower shop"—can be independently verified by multiple parties on the network, such as a domain registrar or a hosting provider. The goal is to create a shared, tamper-resistant source of truth for what an agent is actually allowed to do. Mastercard is initially logging these permissions on Polygon, Solana, and Base.

Multi-Rail Settlement: Cards, Banks, and Stablecoins

AP4M is not a crypto-only product tucked into a corner of the fintech world. It is designed for multi-rail settlement, meaning an organization could use it for traditional card-based payments, bank account transfers, or stablecoin transactions depending on the use case.

• Cards: Transactions route through standard card-network infrastructure.
• Bank accounts: Payments are made via account-based transfers.
• Stablecoins: Settlements can use cryptocurrency rails for value transfer.

This interoperability is a central part of Mastercard's strategy to bridge traditional finance and digital assets, a strategy reflected directly in its choice of launch partners.

The Launch Partners: A Mix of Wall Street and Web3

For its debut, AP4M enlisted more than 30 industry leaders, combining conventional payments giants with crypto-native platforms and cloud infrastructure providers. The roster shows a clear intent to make the service relevant on both sides of the traditional-crypto divide.

Key launch partners include:

• Payments and Fintech: Adyen, Ant International, BVNK, Checkout.com, Getnet by Santander, Global Payments, Stripe, and Tempo.
• Crypto and Blockchain: Coinbase, OKX, Polygon, and the Solana Foundation.
• Infrastructure: Cloudflare and Lovable Labs Incorporated.

Real-World Use Cases for Machine Payments

The most concrete vision for AP4M lies in its use cases, which illustrate how small, programmatic payments can chain together to complete complex tasks. The framework is meant to handle payments that are too small, too fast, or too numerous for human oversight.

• Domain and Web Asset Purchases: An AI agent acting on a single instruction—"set up my online flower shop"—could autonomously purchase a domain name, acquire hosting, and secure related web assets, executing a series of microtransactions across multiple providers.
• Logistics Automation: In a supply chain, an agent could settle freight charges, warehouse fees, and cold-chain monitoring costs in real-time as a shipment progresses, removing the need for batch payments or manual invoicing.
• Per-Use Data Access: An AI agent needing specialized information, such as piecemeal data from a premium API or a paywalled website, could pay a tiny fee for each access, creating a new monetization model for digital content.

The Highly Competitive Race for AI Agent Commerce

AP4M does not enter an empty market. It drops into a rapidly intensifying fight to define the payment architecture for AI agents.

• Visa has been building its own suite of AI payment tools and protocols.
• Coinbase launched its own AI payments protocol directly on the Base blockchain network.
• Stripe, itself an AP4M launch partner, is also pursuing the space with its Machine Payments Protocol in partnership with the blockchain project Tempo.
• Google entered the conversation in September 2025 with the release of its own AI payments standard.

Despite the fanfare, Mastercard’s own leadership has tempered near-term expectations. Jorn Lambert, the company's chief product officer, acknowledged that agentic payment volumes are currently a tiny fraction of total commerce and that AP4M is not projected to be a significant revenue driver in the immediate future. However, Lambert also framed it as a meaningful new addressable market over the next five years, suggesting a belief that AI agents could become intermediaries in a substantial share of e-commerce transactions down the road.