What Is Mastercard Agent Pay for Machines (AP4M)?

What Is Mastercard Agent Pay for Machines?

Mastercard Agent Pay for Machines (AP4M) is a network service, announced June 10, 2026, that allows AI agents and autonomous software to permission, orchestrate, and settle payments at machine speed. It is built for a new class of transactions: micro-value, high-frequency, and executed programmatically between systems rather than initiated by a person tapping a card.

• A new payment class, not a faster card: Mastercard describes AP4M as enabling payments, some only fractions of a cent, that complete quickly, programmatically, and securely between agents that may transact with each other continuously at high velocity.
• Always-on and in the background: unlike point-of-sale or person-to-merchant payments, which are discrete and user-initiated, AP4M transactions are programmatic, always-on, and executed between systems in the background of digital commerce.
• Chains of transactions: a single human request can become a chain of automated payments across providers, including microtransactions, as one agent coordinates and pays for services from many others.
• Network trust extended to machines: AP4M brings the credentialing, controls, and guaranteed settlement of Mastercard's global network to machine-driven commerce, so software can transact with the assurances businesses expect from card payments.

For businesses, the shift is structural rather than incremental. Mastercard frames AP4M as the foundation for a superbloom of AI business models, where any company, from solopreneurs to the largest enterprises, can sell services that agents buy and use. The practical question becomes whether your services are built to be purchased by software, not only by people.

Source: Mastercard Newsroom

Why Does Machine Commerce Need a New Class of Payments?

Machine commerce needs new payment infrastructure because agent transactions break the assumptions traditional payments were built on. Card and person-to-merchant rails assume a human initiates a discrete purchase at a moderate value. Agent payments are continuous, automated, often worth fractions of a cent, and executed between systems with no person in the loop.

• Value too small to process profitably: machine payments can be fractions of a cent, far below the economics that traditional card processing was designed around, so they require a system built for micro-value transactions.
• Frequency too high for human review: agents transact continuously at high velocity, executing chains of payments faster than a person could approve, which calls for authorization rules enforced programmatically rather than per transaction.
• Latency that must approach real time: when one agent pays another for a service it consumes immediately, settlement assurance has to keep pace with execution, or the agent cannot reliably proceed.
• Trust without a human signer: legacy rails authenticate a cardholder, not a piece of software, so machine commerce needs a way to credential an agent and prove it was authorized to act.

Mastercard's framing is that payments in this environment do not simply increase in number, they change form, becoming continuous, embedded, permissioned, and executed at machine speed. Meeting that shift requires infrastructure that can credential software, enforce limits automatically, and guarantee settlement across rails, which is the gap Agent Pay for Machines is built to close.

Source: Mastercard Newsroom

How Does Agent Pay for Machines Actually Work?

Agent Pay for Machines works through four foundational capabilities that together establish a trusted system for machine-driven transactions: credentialing, permissioning, transacting, and settling. Each agent is identified and bounded before it can spend, and every payment settles with a Mastercard guarantee across the rail the recipient prefers.

• Credentialing: every agent is credentialed, and with Verifiable Intent can be recognized and transact with trust across ecosystems, establishing which software is authorized to act in the first place.
• Permissioning: organizations set authorization rules and spending limits that are programmatically enforced, ensuring every transaction stays within defined parameters without a human reviewing each payment.
• Transacting: verified participants connect and transact across providers and systems, enabling continuous, high-frequency automated commerce between counterparties that may never involve a person.
• Settling: AP4M supports reliable, guaranteed multi-rail settlement across cards, accounts, and stablecoins, so transactions move predictably with improved transparency and consistency.

Mastercard's product documentation maps the same model onto a concrete lifecycle, Authorize, Discover, Execute, and Settle, using a coffee shop that assigns an AI agent to build and launch its online presence:

1. Authorize: the agent identifies trusted providers for imagery, content, domain registration, and hosting, then presents a consolidated plan and estimated budget that the business approves with a defined spend limit and preferred funding method.
2. Discover: the approval is recorded in an on-chain smart contract, while network-issued credentials establish trust and a shared discovery layer between the agent and each counterparty.
3. Execute: the agent procures services and completes micro- and macro-payments at machine speed using verifiable vouchers, which are off-chain cryptographic credentials that support high-frequency transactions.
4. Settle: service providers aggregate transactions and submit them for settlement, receiving payout in their preferred currency, whether fiat or stablecoin, backed by a Mastercard settlement guarantee.

Source: Mastercard Agent Pay for Machines

The settlement guarantee is the piece that makes the rest usable. In a chain of automated payments, a business has to know it will get paid even though no human is watching the transaction clear. By backing settlement across cards, accounts, and stablecoins with its global network, Mastercard lets a recipient accept agent-driven payments without taking on the counterparty risk of an unfamiliar agent or an unfamiliar rail.

Source: Mastercard Newsroom

How Is AP4M Different From Mastercard Agent Pay?

Agent Pay for Machines is a complement to Mastercard Agent Pay, not a replacement. Agent Pay, introduced in April 2025, defines how trusted AI agents participate in payments initiated on a person's behalf. AP4M is designed for the opposite end of the spectrum: automated, micro- and machine-driven transactions that run continuously in the background without a human approving each one.

Source: Mastercard Newsroom, Mastercard Agent Pay announcement

Dimension	Mastercard Agent Pay (2025)	Agent Pay for Machines (2026)
Launched	April 2025	June 2026
Who initiates	A trusted agent acting on a person's behalf	Software and agents transacting autonomously
Transaction profile	Discrete, user-initiated purchases	Programmatic, always-on, background
Typical value	Standard consumer purchase amounts	Micro-value, some only fractions of a cent
Volume and speed	Human-paced	Very high volume, very low latency
Settlement rails	Card-network payments	Multi-rail: cards, accounts, stablecoins
Role in the stack	How trusted agents join payments	How machines pay each other continuously

The two programs are designed to stack. Where Agent Pay defines how a trusted agent participates in a user's payment, AP4M scales the high-frequency, low-latency, low-value payments executed by agents and machines once that participation is established. A single business can rely on Agent Pay for consumer-facing purchases and AP4M for the automated, behind-the-scenes settlement between services. The clearest way to see the split is in what AP4M adds on top:

• A payments class for micro-value: Agent Pay handles consumer-scale purchases, while AP4M is engineered for transactions worth fractions of a cent executed at high frequency.
• Background, system-to-system execution: Agent Pay sits inside a user's flow, while AP4M runs continuously between services with no person approving each payment.
• Multi-rail settlement: Agent Pay settles over card-network payments, while AP4M guarantees settlement across cards, accounts, and stablecoins.

What Can Machine Payments Actually Do?

The clearest way to understand AP4M is through the transaction chains it makes possible. Mastercard illustrates machine payments with scenarios where one human instruction triggers many automated payments across independent providers, each too small or too frequent to route through a person.

• Launching a business online: an entrepreneur opening a flower shop instructs an agent to build the store's web presence, buying a domain, hosting, images, and checkout pages within a defined budget, turning one request into a chain of transactions executed across providers.
• Coordinating logistics: a delivery-route agent pays for freight, reserves loading-bay access, purchases temporary cold-chain monitoring data, and settles warehouse handling fees automatically as a shipment moves from origin to destination.
• Paying for compute and resources: partners describe programmatic, machine-led commerce where agents pay for compute, APIs, and data on their own, requiring settlement for the resources they consume in real time.
• Metering agent labor: the business model of many AI agents is metered pricing, which only works once an agent's output can be authorized, metered, and settled with trust, letting agents both pay and get paid.

What unites these examples is volume and velocity. Jorn Lambert, Mastercard's chief product officer, said machine payments make it possible for services to be bought and sold among agents at fundamentally different scales than payments today: very high volumes, very small values, very fast, and at extremely low latency. Payments do not just increase in this model, they change form, becoming continuous, embedded, permissioned, and executed at machine speed.

Launch partners describe the same pattern from the infrastructure side. Cloudflare framed the internet as built for human interactions and argued that agents now need a trusted way to independently pay for the resources they consume. Nevermined described building a Commerce Logic Layer that lets AI agents transact with other agents and agentic services, noting that an agent's labor has to be metered, authorized, and settled with real trust before it can pay or get paid. Together these point to a market where most transactions never involve a person at all.

Source: Mastercard Newsroom

Who Is Building the Machine Payment Economy?

More than 30 companies launched as initial participants and supporters of Agent Pay for Machines, spanning acquirers, stablecoin and blockchain rails, wallet and key-management infrastructure, and agent platforms. The breadth signals that AP4M is positioned as open, multi-rail infrastructure rather than a closed Mastercard product.

Category	Representative launch partners	Role in AP4M
Acquirers and processors	Adyen, Checkout.com, Global Payments, Getnet by Santander, Stripe	Connect merchants and process agent-driven payments
Stablecoin and blockchain rails	Coinbase, BVNK, Ripple, Tempo, Polygon, Solana Foundation	Provide stablecoin settlement and on-chain rails
Wallet and key infrastructure	Turnkey, Crossmint, Utila, Anchorage Digital	Secure agent wallets and digital-asset custody
Agent platforms and trust	Cloudflare, Skyfire, Sapiom, Nevermined, t54 Labs	Build, secure, and verify agents at runtime
Apps and treasury	Lovable, Aave Labs, Alchemy	Agent-built applications and treasury liquidity

Several partners are aligning AP4M with open payment standards they helped build. Tempo, which co-authored the Machine Payments Protocol (MPP) with Stripe in March 2026, said it is working with Mastercard on MPP compatibility, with Tempo providing stablecoin settlement for agent-driven payments at scale.

Source: Mastercard Newsroom, Stripe

Coinbase pointed to x402, the open HTTP-native payment standard it introduced and later moved into a dedicated foundation, as part of an interoperable framework that combines trusted payment networks with programmable digital dollars to enable secure commerce between agents, businesses, and developers.

Source: Mastercard Newsroom, Coinbase

The heavy stablecoin representation among partners is not incidental. Several participants argue that machine payments, happening far more often and at far smaller amounts than today's systems were designed for, are a natural fit for programmable digital dollars that settle in seconds at predictable cost. Mastercard's support for regulated stablecoin settlement on-chain, alongside cards and accounts, lets a recipient accept agent payments in fiat or stablecoin without committing to a single rail.

For Mastercard, the role is the trust and settlement layer beneath these rails. AP4M supports credentialing, controls, and guaranteed settlement across multiple payment types, from cards to stablecoins, so organizations can deploy automated payments with the interoperability, reliability, and governance the digital economy requires, regardless of which underlying rail an agent uses.

Source: Mastercard Newsroom

Where Does AP4M Sit in the Agentic Commerce Stack?

Agent Pay for Machines occupies the settlement layer of the agentic commerce stack, the point where a transaction is permissioned and money actually moves. It sits beneath discovery and execution and works alongside the trust layer Mastercard built with Verifiable Intent. Understanding the full stack clarifies where a business needs to be present to capture agent-driven demand.

• Discovery: before any agent transacts, it has to find and shortlist providers. This is the answer engine optimization layer, where AI assistants such as ChatGPT, Perplexity, and Google AI Overviews decide which businesses enter the consideration set. If an agent cannot discover your service, no payment layer ever reaches you. See the AEO explainer
• Execution: once an agent selects a provider, open commerce protocols govern checkout. The Agentic Commerce Protocol and the Universal Commerce Protocol handle agent-initiated carts for consumer goods, while machine-payment standards handle agent-to-resource purchases for APIs and compute. Compare UCP and ACP
• Trust: Verifiable Intent, the open framework Mastercard developed with Google, creates tamper-resistant proof that an agent was authorized to act. AP4M uses this credentialing so counterparties can recognize and transact with an agent across ecosystems. Read the Verifiable Intent analysis
• Settlement: AP4M is where value clears. It guarantees settlement across cards, accounts, and stablecoins, giving the recipient assurance of payment regardless of how the agent funded the purchase.

Source: Mastercard Newsroom, Mastercard Verifiable Intent

Agent Pay for Machines is the newest layer in a broader Mastercard effort to enable trusted digital interactions, from identity and authentication to trusted data exchange. Together with Agent Pay and Verifiable Intent, it reflects a deliberate sequence: authenticate the agent, prove its intent, then settle its payments. For a business, that sequence doubles as a readiness checklist, because each layer assumes the one before it is already in place.

The strategic implication is that agents are becoming both buyers and sellers, and businesses now have to be discoverable, transactable, and credentialed at once. A company that is invisible to AI discovery never reaches the settlement layer, no matter how capable its payment integration. Readiness starts at the discovery layer and runs through to settlement, which is the lifecycle the A.G.E.N.T.I.C. Framework is built to map.

How Does AP4M Compare to Visa and Other Networks?

Mastercard is not building machine payments alone. Every major card network is extending its rails to AI agents on its own terms. Visa launched Intelligent Commerce in April 2025 and added its Trusted Agent Protocol in October 2025, and American Express shipped its Agentic Commerce Experiences (ACE) Developer Kit in April 2026. Agent Pay for Machines is Mastercard's entry in that same race.

Source: Visa Intelligent Commerce, Visa Trusted Agent Protocol, American Express

Network	Agentic framework(s)	First launched
Mastercard	Agent Pay, Agent Pay for Machines, Verifiable Intent	April 2025 (Agent Pay)
Visa	Visa Intelligent Commerce, Trusted Agent Protocol	April 2025 (Intelligent Commerce)
American Express	Agentic Commerce Experiences (ACE) Developer Kit	April 2026

The networks compete on frameworks but cooperate on protocols. Agent Pay for Machines, Visa Intelligent Commerce, and Amex ACE are each governed by a single network. They sit alongside the cross-network protocols that move between rails: AP2, Google's Agent Payments Protocol, plus the Machine Payments Protocol from Stripe and Tempo, and x402 from Coinbase. Treating a network framework and a cross-network protocol as the same thing is a common mistake. See the Visa Intelligent Commerce analysis and the Amex ACE breakdown.

Mastercard's relationship with the open protocols is collaborative rather than competitive. Mastercard is one of more than 60 organizations Google named as Agent Payments Protocol collaborators when AP2 launched in September 2025, and Verifiable Intent, the credentialing layer Agent Pay for Machines relies on, was built with Google to be compatible with AP2. For a business the convergence is good news: the same discoverability and structured data that make you transactable through Agent Pay for Machines also position you for Visa and Amex agent flows, because every one of them depends first on an agent being able to find you. Read the Verifiable Intent analysis

Source: Google Cloud, Mastercard Verifiable Intent

The businesses that act early will not be integrating a single product so much as positioning themselves across an emerging stack. Machine payments reward the companies that are already findable by AI, already structured for agent checkout, and already prepared to settle across whatever rail an agent brings. The readiness path below moves a business from invisible to transactable in an agent-led economy. See how agentic commerce is taking shape