When machines start making payments, Mastercard aims to rewrite the entry point for digital commerce.
Written by: KarenZ, Foresight News
On June 10, Mastercard launched Agent Pay for Machines, aiming to enable high-speed, low-value, continuous payments between machines in the backend.
This isn't just about adding another payment button for regular consumers; it's a settlement infrastructure layer prepared for enterprise systems, API services, logistics nodes, and data service providers' AI Agents.
What's interesting about this whole Web3 thing is that Mastercard isn't treating stablecoins as some fringe crypto payment experiment. They're listing it alongside cards and accounts as part of their Agent Pay for Machines settlement track.
In other words, stablecoins are being absorbed into traditional payment networks using a more familiar business language: identity, authorization, limits, audits, and settlement guarantees.
How does Agent Pay for Machines operate?
Agent Pay for Machines is built on the foundation of Mastercard Agent Pay launched in 2025. The latter addresses how AI Agents can participate in payments with a trusted identity, while the former further targets machine-to-machine scenarios, handling higher frequency, lower latency, smaller, and continuously occurring backend transactions.
Mastercard's framework for Mastercard Agent Pay for Machines can be broken down into four layers: authentication, permissions, transactions, and settlements.
The first layer is authentication. Each Agent needs to be credentialed, meaning they must first obtain a recognizable and traceable identity. Mastercard emphasizes 'Know your agent' on the Agent Pay page, stating that only registered Agents can transact.
Verifiable Intent is also a key piece of this authentication system. Mastercard and Google jointly launched this open, standards-based trust layer in March 2026, stating that it aligns with Google's Agent Payments Protocol and Universal Commerce Protocol while remaining protocol-agnostic. The problem it aims to solve is very specific: when an AI Agent acts on behalf of a user, how do external parties know that this transaction aligns with the user's original authorization?
Its approach is to link identity, intent, and actions into a verifiable record: who authorized the Agent, what the authorization entailed, and what interactions occurred between the Agent and the merchant (leading to this purchasing behavior). In case of disputes, all parties can rely on the same audit trail to determine liability. For consumers, delegation does not equal loss of control; for merchants and issuers, transactions initiated by the Agent are no longer just a black box request.
The second layer is permissions. Enterprises can set authorization rules and spending limits for Agents and have these rules executed programmatically.
For example, a procurement Agent may be allowed to purchase cloud computing services up to a certain amount but cannot use the company credit line to buy unrelated assets; a logistics Agent can automatically settle storage fees, but if it exceeds the budget or takes an unusual route, it must request confirmation again.
The third layer is transactions. Verified participants can connect across service providers and systems to complete continuous, high-frequency, automated transactions. This stage is the truest entry point to the 'machine economy': Agents are no longer just clicking buttons on a webpage but can coordinate services, purchase resources, and execute tasks across multiple providers.
The fourth layer is settlement. Mastercard states that Agent Pay for Machines supports various payment types, including cards, accounts, and stablecoins, providing reliable, deterministic multi-track settlements. Stablecoins are embedded at this layer: they do not solely bear the entire payment experience but are included as one of the settlement tracks within Mastercard's familiar network, rules, and governance framework.
This also explains why Mastercard is incorporating stablecoins into Agent Pay for Machines instead of only discussing credit cards. Machine payments are naturally suited for micro, high-frequency, programmable settlements, and stablecoins have advantages in speed, cost, and programmability; however, enterprise-level adoption still requires identity, permissions, risk control, refunds, dispute resolution, and liability assignment. What Mastercard aims to do is embed the settlement efficiency of stablecoins into a payment control system that can be accepted by merchants, institutions, and enterprises.
The collaboration list is a new payment landscape.
Mastercard announced that over 30 initial participants and supporters are involved, including: Aave Labs, Adyen, Alchemy, Anchorage Digital, Ant International, Basis Theory, BVNK, Catena, Checkout.com, Cloudflare, Coinbase, Coinflow, Crossmint, Getnet by Santander, Global Payments, Lovable, Mastercard Merchant Cloud, MoonPay, Nevermined, OKX, PayOS, Polygon, Rain, Ripple, Sapiom, Skyfire, Solana Foundation, Stripe, t54 Labs, Tempo, Turnkey, and Utila.
This list can be divided into several categories of players.
The first category is traditional payments, merchant services, and fintech networks, including Adyen, Ant International, Checkout.com, Getnet by Santander, Global Payments, Stripe, Mastercard Merchant Cloud, and others. They connect merchants, acquirers, processors, and checkout experiences, serving as the entry point for Agent payments into real business scenarios.
The second category includes stablecoin infrastructure, exchanges, public chains, and on-chain liquidity networks. Among them, Coinbase aims to work with Mastercard to promote an open, interoperable Agent payment framework, combining trusted payment networks, programmable digital dollars, and open standards like x402, enabling secure and scalable transactions between Agents, enterprises, and developers. OKX will connect to this ecosystem through Agentic Wallet and the Agent Payments Protocol (APP).
Stripe and Paradigm's incubated payment public chain Tempo explicitly mentions providing stablecoin settlements for large-scale Agent payments. RippleX has integrated XRPL and RLUSD into enterprise Agent payment scenarios. Other public chains include Solana, Polygon, and more.
Aave Labs' role is more on the funding layer, providing foundational credit layers and deep liquidity for AI agents to borrow and earn.
The third category is the intermediate layer for Agent payments, including Alchemy, Cloudflare, Lovable, Nevermined, PayOS, Sapiom, Skyfire, t54 Labs, Catena, Basis Theory, Crossmint, and others. They fill roles related to developer tools, network connectivity, identity and credentials, payment authorization, transaction risk control, inter-agent transaction orchestration, and the commercial logic between Agents and merchants.
The focus of this list is not on 'who stands on stage' or 'who supports', but on the different gaps each type of player fills. Merchant service providers handle integration, cloud and developer platforms provide the Agent operating environment, wallets and custodians manage keys and asset control, public chains and stablecoin networks handle settlements, DeFi offers liquidity and credit layers, while risk control and identity companies manage authorization, audits, and liability.
If in the past, Web3 payments were often framed as alternatives to bypass traditional networks, this time it resembles an interface convergence: card networks, account systems, stablecoins, public chains, acquiring institutions, and Agent platforms are all lumped together to compete for the payment standards of the machine economy.
Stablecoins entering machine payments must first learn to be constrained.
Agent payments are the easiest to spin into efficiency tales: machines are faster than humans, micro-payments are more flexible than subscriptions, and usage-based billing is fairer than monthly fees. But as the Agent moves from 'generating suggestions' to 'executing payments', risks extend from information errors to authorization breaches, fund misuse, and liability issues.
This is precisely why verifiable intent, Know your agent, permission rules, and multi-track settlements keep coming up. In the future, payment networks will need to recognize not just a card, an account, or a wallet address, but a software entity representing user or enterprise actions. It must know who this Agent is, what they are permitted to do, if the money aligns with the original intent, and who is responsible if issues arise.
For Web3, this means a challenging reality: for stablecoins to enter enterprise-level Agent payments, it’s not enough to just have on-chain transfer speeds. Enterprise clients will demand limits, approvals, audits, compliance, dispute resolution, and accountability. Without these controls, micro-payments could turn into a wholesale market of micro incidents; if controls are too stringent, the machine economy may revert back to a few closed platforms.
So the signal emitted by Agent Pay for Machines resembles a translation of stablecoins into a language understood by institutions: it can be a settlement track, a programmable funding layer, or serve high-frequency trading between Agents, but it must fit into a verifiable, authorized, and governable business framework.
The next generation of payment gateways may not appear at the checkout counter but between API requests, logistics nodes, cloud service calls, Agent task distributions, and backend settlements. Humans provide an intent, machines break it down into a series of transactions, and the underlying settlement can choose between cards, accounts, or stablecoins as tracks.
Stablecoins finally have a real chance to become machine-readable business fuel. But to enter the mainstream, they may first need to learn to run within the rules.