Mastercard has launched Agent Pay for Machines (AP4M), a new payments infrastructure layer designed to allow AI agents to execute, permission, and settle transactions at machine speed across its global network.
The service, announced June 10, is the card network's most direct move yet into the emerging category of autonomous machine-to-machine commerce.
What AP4M does
Traditional payment rails are built for human-initiated transactions — one purchase, one confirmation. AP4M is built for a different model: continuous, programmatic payments between AI agents at sub-cent values that conventional processing can't handle profitably.
Mastercard says the system can process fractions of a cent per transaction, removing the minimum-economically-viable-transaction-size constraint that has made high-frequency agent payments impractical until now.
The service includes credentialing and permissioning features that let an AI agent prove it is authorised to act on behalf of a user, and lets users set programmable rules and spending limits that are enforced automatically. A user could instruct an AI agent to build a website and have that agent independently purchase domain, hosting, imagery, and payment processing across multiple providers — all within a defined budget, without human confirmation at each step.
"Machine payments can make it possible for services to be bought and sold among agents at fundamentally different scales than payments today," said Jorn Lambert, Mastercard's chief product officer. "Very high volumes, very small values, very fast and at extremely low latency."
Built on public blockchains; crypto and fintech partners
A notable design choice: Mastercard is recording permissions granted to AI agents on public blockchains rather than in a private database. The initial chains selected are Polygon, Solana, and Base, allowing multiple parties to independently verify whether an agent is operating within its authorised scope without relying on any single company's records.
Polygon said it would "help establish common rules for agent-to-agent use cases and accelerate the adoption of agentic commerce."
More than 30 companies joined as early partners. Crypto-native participants include Coinbase, OKX, Ripple, Solana, Polygon, Aave Labs, Anchorage Digital, BVNK, MoonPay, and Alchemy. Broader fintech and cloud participants include Stripe, Cloudflare, Checkout.com, and Ant International. Tempo and Coinbase are separately developing open payment protocols — the Machine Payments Protocol and x402 — aimed at standardising how agents authorise and route payments across providers.
Mastercard's stablecoin and crypto rails
AP4M sits within Mastercard's broader expansion of crypto-compatible infrastructure. The company already supports stablecoin-based card settlement for USDC, PYUSD, and RLUSD, and is running an initiative with more than 100 partners to integrate blockchain and stablecoin flows with traditional rails for cross-border transfers, B2B payments, and commerce. Mastercard acquired stablecoin startup BVNK in March and holds a New York State BitLicense.
AP4M's significance is threefold. It creates an institutional-grade on-ramp for stablecoin micropayments at scale — a long-standing infrastructure gap. It validates on-chain permissioning as a shared verification model for machine authorisation. And by embedding this into a global card network processing trillions annually, Mastercard is normalising AI agent payment capabilities within existing financial infrastructure rather than requiring agents to operate solely within crypto-native ecosystems.
The initial phase will focus on validating use cases with launch partners. No broader rollout timeline or fee schedule disclosed.
