Mastercard launched Agent Pay for Machines (AP4M) on June 10, 2026, an open protocol enabling AI agents to authorize, coordinate, and settle transactions across cards, bank accounts, and stablecoins on its global network. More than 30 companies joined as launch partners, including Coinbase, OKX, ...
Mastercard launched Agent Pay for Machines (AP4M) on June 10, 2026, an open protocol enabling AI agents to authorize, coordinate, and settle transactions across cards, bank accounts, and stablecoins on its global network. More than 30 companies joined as launch partners, including Coinbase, OKX, Stripe, Aave Labs, Cloudflare, and the Solana Foundation. Agent credentials will be recorded on Polygon, Solana, and Base blockchains, with broader chain support planned for later in 2026.
The announcement places Mastercard in direct competition with at least four other infrastructure standards for machine payments: Coinbase's x402 protocol, Stripe's Machine Payments Protocol (MPP) on the Tempo blockchain, Google's AP2 delegated-spending framework, and Visa's tokenized-credential system for AI-driven commerce. The combined market for agentic commerce is projected between $1.5 trillion (Juniper Research) and $3–5 trillion (McKinsey) by 2030. Who controls the settlement layer for machine-to-machine transactions will determine a significant share of payments economics for the next decade.
AP4M adds a network layer on top of Mastercard's existing payments infrastructure, specifically targeting high-volume, low-value transactions that legacy rails handle poorly or not at all. The protocol provides three core functions:
The system integrates with HTTP 402, the long-dormant "Payment Required" status code now being revived as an internet-native payment standard. According to Raj Dhamodharan, Mastercard's executive vice president of blockchain and digital asset products: "There are already many declines happening because there is no payment option available. That is a leading indicator."
The use cases span API access billing, cloud compute procurement by autonomous agents, AI-to-AI service marketplaces, and machine-initiated micropayments for data feeds. These transactions typically range from fractions of a cent to a few dollars — volumes where card network minimum fees make traditional rails economically unviable.
AP4M records agent permissions and credentials on three blockchains at launch: Polygon, Solana, and Base (Coinbase's Ethereum L2). The choice reflects a pragmatic calculation. All three offer sub-second finality and transaction costs measured in fractions of a cent, making them suitable for the micropayment volumes AP4M targets.
The on-chain component handles identity and authorization. Settlement itself remains flexible — transactions can close on Mastercard's traditional rails (card networks, ACH) or on-chain via stablecoins. This hybrid architecture avoids forcing participants to choose between fiat and crypto infrastructure.
Mastercard has not disclosed specific throughput targets for AP4M, but the system is designed for "machine speed" — implying sub-second authorization and settlement latency, a requirement for autonomous agents operating in real time.
The 30+ launch partners span five categories:
| Category | Partners | |---|---| | Crypto exchanges | Coinbase, OKX | | DeFi protocols | Aave Labs | | Blockchain platforms | Polygon, Solana Foundation, Alchemy | | Payment processors | Stripe, Adyen, Checkout.com, Global Payments, Getnet (Santander) | | Infrastructure | Cloudflare, Anchorage Digital, BVNK, MoonPay, Ripple, Turnkey, Utila |
Nathan McCauley, CEO of Anchorage Digital, described the system as combining "Mastercard's trust and global reach with multi-rail settlement flexibility, including digital assets." Stephanie Cohen, Cloudflare's chief strategy officer, positioned it as connecting "developer and security platform with world-class payments infrastructure."
The breadth of the partner list is notable. Having both Coinbase (which operates the competing x402 protocol) and Stripe (which launched its own Machine Payments Protocol) as AP4M partners suggests these companies view protocol interoperability, not exclusivity, as the near-term strategy.
AP4M enters a market where at least four competing standards are already live or in advanced testing:
Coinbase x402: Launched as an open protocol reviving HTTP 402, x402 has reached 69,000 active AI agents, 165 million transactions, and $50 million in cumulative volume as of April 2026. Base and Solana are its primary settlement chains. The x402 Foundation, co-founded by Coinbase and Cloudflare, counts Google, Visa, AWS, Circle, Anthropic, and Vercel among its core members.
Stripe Machine Payments Protocol (MPP): Stripe built MPP on its Tempo blockchain, positioning it as a native settlement layer for agent commerce. Details on transaction volume remain limited.
Google AP2: Google's delegated-spending authorization system allows users to grant AI agents scoped payment permissions. Integration with Google's cloud and AI infrastructure gives it a distribution advantage within the Google ecosystem.
Visa Intelligent Commerce: Visa launched Intelligent Commerce Connect, a platform allowing AI agents to make payments across multiple card networks. Visa has also announced a strategic collaboration with OpenAI to support payments inside agentic commerce experiences.
Amazon Bedrock AgentCore Payments: AWS launched agent payment infrastructure in May 2026, built with Coinbase and Stripe, enabling AI agents on Amazon Bedrock to transact using stablecoins.
The emerging picture is fragmentation. No single protocol has achieved dominant market share, and several major players are hedging by participating in multiple standards simultaneously.
Market size estimates for agentic commerce vary by an order of magnitude depending on scope and methodology:
| Source | Projection (by 2030) | Scope | |---|---|---| | McKinsey | $3–5 trillion | Global agentic commerce revenue | | Gartner | $15 trillion by 2028 | AI agent-intermediated purchases | | Juniper Research | $1.5 trillion | Global agentic commerce spending (61 countries, 38,000 data points) | | Morgan Stanley | $190–385 billion | U.S. e-commerce only |
The specialized agentic payment infrastructure market — the billing and settlement layer itself — is projected to grow from $7 billion to $93 billion by 2032, according to industry estimates cited by Nevermined.
These figures carry significant uncertainty. Each institution defines "agentic commerce" differently, and the market depends on AI agent capabilities that are still evolving rapidly. What the projections agree on is directionality: machine-initiated transactions are growing from negligible to material.
Between May 2025 and April 2026, AI agents settled more than $73 million across approximately 176 million blockchain transactions, according to a Keyrock report cited by CoinDesk. USDC accounted for 98.6% of settlement volume.
The economics explain the concentration. Approximately 76% of agent transactions fall below the 30-cent fixed-fee floor common in card payments. Most payments range between 1 and 10 cents. At these values, card network economics break down — a $0.30 minimum fee on a $0.05 transaction represents a 600% surcharge. Stablecoin settlement on chains like Base and Solana costs fractions of a cent, making sub-dollar transactions viable.
This creates a structural advantage for stablecoin-based settlement in the agent economy. Traditional payment networks can compete on trust, compliance, and existing merchant relationships, but their fee structures were built for human-scale transaction values. AP4M's multi-rail design acknowledges this reality by including stablecoins alongside cards and bank accounts.
The near-total dominance of USDC introduces concentration risk. Circle's stablecoin serves as effectively the sole settlement currency for machine payments. Any disruption to USDC — regulatory, operational, or competitive — would affect the entire agent payment ecosystem.
AP4M is the latest in a series of Mastercard moves into digital asset infrastructure:
The BVNK acquisition is particularly relevant. BVNK's infrastructure connects on-chain payment flows with fiat rails — precisely the bridge AP4M requires for multi-rail settlement. Integrating BVNK's technology gives Mastercard a proprietary stablecoin on/off-ramp within its agent payment stack.
Protocol fragmentation: Five competing standards with overlapping partner lists suggests the market may consolidate around two or three protocols, but the timeline is unclear. Enterprises building on one protocol face switching costs if a different standard wins.
Regulatory ambiguity: The U.S. regulatory framework for AI agent transactions remains undefined. California's Digital Financial Assets Law takes effect July 1, 2026, but it was not designed with autonomous agent commerce in mind. Questions about liability — who is responsible when an AI agent makes an unauthorized purchase — remain unresolved.
Consumer trust deficit: According to data compiled by Nevermined, only 16% of U.S. consumers and 29% of U.K. consumers trust AI-made payments. Adoption at scale requires bridging this gap, and early fraud incidents could set the market back.
USDC concentration: With 98.6% of agent settlement in a single stablecoin, systemic risk remains elevated. Multi-stablecoin settlement support exists in theory but not yet in practice.
Valuation uncertainty: The gap between McKinsey's $3–5 trillion and Morgan Stanley's $190–385 billion projections illustrates how loosely defined the market remains. Investment decisions based on the high-end estimates carry proportional risk.
Mastercard's AP4M announcement marks the point where machine-to-machine payments moved from experimental infrastructure to a product backed by one of the world's two largest card networks. The protocol's hybrid design — blending blockchain-based credentialing with traditional and stablecoin settlement — reflects a pragmatic read of a market where neither pure crypto nor pure fiat rails alone serve the economics of sub-cent transactions.
The immediate question is not whether agent payments become a significant market, but which settlement standard captures the largest share. Mastercard is betting that trust, compliance infrastructure, and existing merchant relationships give it an advantage over crypto-native alternatives. Coinbase, Stripe, and Google are making the opposite bet — that developer adoption and protocol openness matter more than legacy network effects.
Both sides may be right. The market is large enough to support multiple winners, and the 30+ partners that simultaneously belong to AP4M, x402, and other standards suggest the industry itself expects coexistence rather than winner-take-all. The economic value will accrue to whoever controls settlement finality — the moment a transaction becomes irreversible. That is the layer Mastercard, with its BVNK acquisition and AP4M protocol, is positioning to own.