Mastercard launches Agent Pay for Machines, an open payments protocol for AI agents with microtransaction support and 30+ partners including Coinbase, Stripe, and Adyen

Mastercard launched Agent Pay for Machines on June 10, 2026 — an open payments protocol designed to let AI agents and autonomous systems execute high-frequency, low-value transactions programmatically, without human intervention. More than 30 companies have joined the ecosystem at launch, including Coinbase, Stripe, Adyen, Cloudflare, OKX, Ant International, and Checkout.com.

What's new

Agent Pay for Machines (AP4M) addresses a structural gap: existing payment infrastructure was built for discrete, human-initiated transactions. AI agents operating on behalf of users need a different model — one where spending is governed by machine-readable policies and transactions execute continuously in the background.

The protocol is organized around four capabilities:

• Credentialing: AI agents receive a verified identity through a "Verifiable Intent" mechanism, allowing multiple parties to confirm the agent is authorized before a transaction settles
• Permissioning: Human-granted spending limits and category restrictions are encoded as programmable authorization rules that enforce automatically — no human needs to approve each transaction
• Transacting: Verified agents connect across Mastercard's partner network for always-on commerce
• Settlement: Multi-rail support across cards, bank accounts, and stablecoins (with permissions stored on Polygon, Solana, and Base)

The protocol supports micropayments worth fractions of a cent — enabling agents to pay for API calls, data, or compute at granularities that traditional card rails were not optimized for.

Mastercard chief product officer Jorn Lambert framed the commercial outlook: "Do I think it'll be a meaningful new addressable market for us over the next five years? I think so."

Context

Visa entered adjacent territory in a May 2026 deal with OpenAI enabling ChatGPT to handle checkout flows. Mastercard's AP4M is broader in scope: it is an open protocol with interoperability across multiple payment rails and a 30+ partner ecosystem at launch, designed for agent-to-agent and agent-to-service transactions rather than consumer checkout.

The blockchain credential layer is a meaningful architectural choice. Storing agent permissions on public chains (Polygon, Solana, Base) means any verifying party in the ecosystem can check whether a given agent is authorized to transact without depending on a centralized Mastercard registry.

The use case Mastercard highlighted at launch: an entrepreneur tasks an agent with building a business's digital presence. The agent independently purchases a domain, hosting, marketing materials, and payment tools — all within a pre-set budget — without requiring the entrepreneur to approve each transaction individually.

Why it matters

For AI labs building agents, payment capability is increasingly table stakes. Agents that can only retrieve information are limited; agents that can execute transactions unlock a much larger set of useful tasks — booking, procurement, logistics, API consumption. AP4M gives any AI system a governed, auditable way to spend money on behalf of a user.

The 30+ partner roster spans crypto infrastructure (Coinbase, Anchorage Digital, Alchemy), traditional payment processors (Adyen, Checkout.com, Global Payments), and developer infrastructure (Cloudflare, Lovable Labs). That breadth gives AP4M meaningful coverage from day one.

Mastercard is effectively betting that machine-to-machine commerce will grow into a substantial new payment category over the next five years — and that whoever owns the agent credential and settlement layer will be well-positioned in that market.