# Mastercard Agent Pay for Machines (AP4M)

> A Mastercard network service for high-frequency, low-latency machine-to-machine payments. Builds on the Agent Pay program (2025) with credentialing via Verifiable Intent, programmable permissioning, and multi-rail settlement. Launched June 2026 with 30+ partners including Stripe, Adyen, Coinbase, and Cloudflare.

## At a glance

| Field | Value |
| --- | --- |
| Name | AP4M |
| Layer | payment-execution |
| Category | card-rail |
| Status | live |
| Created by | Mastercard |
| Launched | 2026-06 |
| Rails | cards, bank accounts, stablecoins |
| Settlement | Multi-rail settlement across cards, bank accounts, and stablecoins via Mastercard's global network. |
| Reversible | yes |
| Human in loop | optional |
| D2C fit | moderate |
| B2B fit | strong |
| Spec | https://www.mastercard.com/us/en/news-and-trends/press/2026/june/mastercard-launches-agent-pay-for-machines.html |
| Last verified | 2026-09-14 |

## How it works

AP4M extends Mastercard's **Agent Pay** program (2025) from human-delegated agent transactions to **fully autonomous machine-to-machine commerce** — payments that run continuously in the background with no human in the transaction loop.

- **Credentialing.** Every agent is credentialed on Mastercard's network. **Verifiable Intent** (co-developed with Google and donated to FIDO alongside AP2) provides a tamper-proof record of what the agent is authorized to do, so counterparties can cryptographically verify an agent's identity and mandate.
- **Permissioning.** Organizations set authorization rules and spending limits that are **programmatically enforced** — per-transaction caps, per-period budgets, and allowed merchant categories — ensuring transactions stay within defined parameters without per-transaction human approval.
- **Transacting.** Verified agents connect and transact across providers and systems, enabling continuous, high-frequency automated commerce. Mastercard's network handles routing and orchestration at machine speed.
- **Settling.** Multi-rail settlement supports **cards, bank accounts, and stablecoins**. Recipients receive payouts in their preferred currency, and the settlement layer provides the guarantee and finality that machine-speed commerce requires.

AP4M is positioned for use cases like an AI agent autonomously purchasing compute, hosting, and checkout infrastructure, or a logistics agent paying freight and monitoring fees — transactions that are high-frequency, low-value, and fully automated.

## Tradeoffs

- Operates within Mastercard's centralised network and credentialing model, which is a trust advantage for regulated commerce but limits permissionless participation.
- Primary Mastercard documentation URLs return 403 to automated fetchers; details here are verified against the investor relations announcement and third-party coverage.
- As a newly launched service (June 2026), production-scale transaction volumes and edge cases are still being validated with the 30+ launch partners.
- Multi-rail settlement (cards + accounts + stablecoins) adds flexibility but also complexity in reconciliation and dispute handling across different rails.

## Security & limitations

- Agent credentialing and Verifiable Intent are load-bearing trust primitives; a compromised credential could authorize unauthorized machine-speed spending.
- Programmatic spending limits must be enforced atomically and in real time to prevent race conditions or burst-spending exploits at machine speed.
- The multi-rail settlement model means security properties differ per rail — card disputes, bank reversals, and stablecoin finality each have different risk profiles.

## Fits together with

- ap2
- x402

## Sources

1. [Mastercard launches AP4M for AI-led transactions](https://www.electronicpaymentsinternational.com/news/mastercard-launches-ap4m-for-ai-led-transactions/) — Electronic Payments International, 2026-06-11
2. [Mastercard launches Agent Connect for agentic commerce](https://thepaypers.com/payments/news/mastercard-launches-agent-connect-for-agentic-commerce) — The Paypers, 2026-09-10

---
_Source: payful.ai — independent editorial reference. Verified 2026-09-14._