When machines start to pay, Mastercard wants to rewrite the digital commerce entry.

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 background.

It is not just an additional payment button for regular consumers, but a layer of settlement infrastructure prepared for business systems, API services, logistics nodes, and data service providers' AI Agents.

The interesting aspect of this for Web3 is that Mastercard does not treat stablecoins as a marginalized experimental line of crypto payments, but lists them alongside cards and accounts as a supported settlement track for Agent Pay for Machines.

In other words, stablecoins are being absorbed into a more familiar commercial language by traditional payment networks: identity, authorization, limits, audit, settlement guarantees.

How does Agent Pay for Machines work?

The base of Agent Pay for Machines comes from Mastercard's Mastercard Agent Pay launched in 2025. The latter addresses how AI Agents can participate in payments with trusted identities, while the former further targets machine-to-machine scenarios, handling higher frequency, lower latency, smaller, continuously occurring backend transactions.

The operating framework provided by Mastercard for Mastercard Agent Pay for Machines can be broken down into four layers: authentication, permissions, transactions, and settlement.

The first layer is authentication. Each Agent needs to be credentialed, meaning it must gain a recognizable, 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, standard-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 specific problem it addresses is: 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 connect identity, intent, and action into a verifiable record: who authorized the Agent, what the authorization content was, and what interactions occurred between the Agent and the merchant (leading to the purchase behavior). In case of disputes, parties can refer to the same audit trail to determine responsibility. For consumers, delegation does not mean loss of control; for merchants and issuers, transactions initiated by Agents are no longer just black box requests.

The second layer is permissions. Businesses 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 company credit for unrelated assets; a logistics Agent can automatically settle warehouse fees but must re-request confirmation if the budget is exceeded or an unusual route occurs.

The third layer is transactions. Verified participants can connect across service providers and systems to complete continuous, high-frequency, automated transactions. This phase closely resembles the true entry into the "machine economy": Agents can coordinate services, purchase resources, and execute tasks among multiple vendors instead of just clicking buttons on web pages.

The fourth layer is settlement. Mastercard states that Agent Pay for Machines supports various payment types such as cards, accounts, and stablecoins, providing reliable, deterministic multi-track settlement. Stablecoins are embedded at this layer: they do not solely bear the entire payment experience but serve as one of the settlement tracks within Mastercard's familiar networks, rules, and governance framework.

This also explains why Mastercard includes stablecoins in Agent Pay for Machines, rather than just discussing bank cards. Machine payments are inherently suited for small, high-frequency, programmable settlements, with stablecoins having advantages in speed, cost, and programmability; however, enterprise-level adoption still requires identity, permissions, risk control, refunds, dispute resolution, and accountability. What Mastercard aims to do is incorporate the settlement efficiency of stablecoins into a payment control system that can be accepted by merchants, institutions, and enterprises.

The partner list: a new payment landscape

Mastercard has announced that more than 30 initial participants and supporters are involved, with a complete list 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 includes traditional payment, merchant services, and fintech networks, such as Adyen, Ant International, Checkout.com, Getnet by Santander, Global Payments, Stripe, and Mastercard Merchant Cloud. They connect merchants, acquiring, processing, and checkout experiences, serving as the entry point for Agent payments into real commercial scenarios.

The second category includes stablecoin infrastructure, exchanges, public blockchains, and on-chain liquidity networks. Among these, Coinbase hopes to work with Mastercard to promote an open, interoperable Agent payment framework that combines trusted payment networks, programmable digital dollars, and open standards like x402, enabling secure and scalable transactions among Agents, enterprises, and developers. OKX will connect to the ecosystem through Agentic Wallet and the Agent Payments Protocol (APP).

Stripe and the payment public blockchain Tempo incubated by Paradigm explicitly mention providing stablecoin settlement for large-scale Agent payments. RippleX incorporates XRPL and RLUSD into enterprise Agent payment scenarios. Other public blockchains include Solana, Polygon, etc.

Aave Labs plays a more funding-oriented role, providing a foundational credit layer and deep liquidity for AI agents to borrow and earn.

The third category consists of middle-layer solutions for Agent payments, including Alchemy, Cloudflare, Lovable, Nevermined, PayOS, Sapiom, Skyfire, t54 Labs, Catena, Basis Theory, Crossmint, etc. They fill the gaps: developer tools, network connections, identity and credentials, payment authorization, transaction risk control, inter-Agent transaction orchestration, and business logic between Agents and merchants.

The emphasis of this list is not on "who stands on the platform" or "who supports," but on how each type of player addresses different gaps. Merchant service providers tackle integration, cloud and developer platforms solve the Agent operating environment, wallets and custody manage keys and asset control, public blockchains and stablecoin networks handle settlements, DeFi offers liquidity and credit layers, while risk control and identity companies address authorization, audits, and accountability.

If past discussions on Web3 payments often portrayed them as alternative solutions bypassing traditional networks, this time it resembles a convergence of interfaces: card networks, account systems, stablecoins, public blockchains, acquiring institutions, and Agent platforms are all placed within the same question, competing for standards in the machine economy.

Stablecoins entering machine payments must first learn to be constrained

Agent payments are easily framed as efficiency stories: machines are faster than humans, micropayments are more flexible than subscriptions, and usage-based billing is fairer than flat-rate monthly fees. However, as Agents transition from "generating recommendations" to "executing payments," risks can expand from misinformation to authorization overreach, misuse of funds, and accountability issues.

This is precisely why verifiable intent, Know your agent, permission rules, and multi-track settlement repeatedly emerge. The future payment network must recognize entities that are no longer just a card, account, or wallet address, but rather a software entity representing user or enterprise actions. It must understand who this Agent is, what they are permitted to do, whether the expenditure aligns with the original intent, and who is responsible if issues arise.

For Web3, this signifies a challenging reality: for stablecoins to enter enterprise-level Agent payments, simply having on-chain transfer speeds is insufficient. Corporate clients will also demand limits, approvals, audits, compliance, dispute resolution, and accountability. Without these controls, micropayments may turn into a wholesale market for minor incidents; if controls are too stringent, the machine economy may revert to a few closed platforms.

Therefore, the signal released by Agent Pay for Machines resembles a retranslation of stablecoins into a language comprehensible to institutions: they can serve as settlement tracks, as programmable funding layers, and support high-frequency trading between Agents, but must fit into a verifiable, authorizable, and governable business framework.

The next-generation payment entry may not appear at the checkout counter but rather among API requests, logistics nodes, cloud service calls, Agent task distributions, and backend settlements. Humans provide an intent, and machines break it down into a series of transactions, with underlying settlements able to choose from card, account, or stablecoin tracks.

Stablecoins truly have the opportunity to become machine-readable commercial fuel for the first time. However, to enter the mainstream, they may first need to learn to run within rules.