Mastercard Launches AP4M Protocol for AI Agent Payments and Stablecoin Settlement

Agentic payments allow AI systems to compare products, place orders and complete transactions within limits set by users. The model could reshape digital commerce but raises unresolved questions around agent identity, proof of consumer intent, fraud liability and interoperability across card and blockchain networks. Rain, Visa and Mastercard are seeking common rules to prevent the emerging market from fragmenting into incompatible, proprietary systems.

Rain has formed the Agentic Payments Alliance, bringing together Visa, Mastercard, Circle, Solana and other payments and blockchain participants. The group plans to develop technical and governance guidance for AI commerce and autonomous-agent payments. It has not disclosed a funding commitment, implementation date or completed standard, leaving the initial effort focused on a shared framework spanning traditional card networks, stablecoins and blockchain settlement infrastructure.

Generative AI is evolving from answering questions to shopping, subscribing to services and initiating transactions on users’ behalf. That shift exposes a gap in payment infrastructure built around human cardholders and conventional merchant checkouts. Visa and Mastercard are developing dedicated rails for AI agents, while stablecoin and cryptocurrency protocols are targeting micropayments, machine-to-machine settlement and the identity and authorization controls autonomous commerce requires.

The latest reported tally as of Aug. 6, 2026, put AI-agent payment volume at only about $136 million, underscoring how early the market remains. Visa and Mastercard are competing to control the gateway for automated transactions, while onchain providers are pitching programmable, low-cost stablecoin payments for small purchases. Both camps are seeking to solve a central constraint: enabling an AI agent to prove its identity and spending authority before money moves.

Visa, Mastercard, American Express and Discover are adapting card networks for a “Prompt Economy” in which AI agents move from product discovery to purchasing on a user’s behalf. The networks are building tokenized credentials, agent identity checks and verifiable mandates into existing rails, seeking to preserve fraud controls, refunds and dispute rights when software initiates a transaction. The shift matters because agents could ultimately choose both merchants and payment methods, threatening to bypass incumbent networks through stablecoins and open internet protocols.

Mastercard on June 10, 2026, launched Agent Pay for Machines, enabling software-driven payments to be permissioned, orchestrated and settled at machine speed across cards, bank accounts and stablecoins. More than 30 companies backed the initial rollout, and transactions can be worth fractions of a cent. The move followed Mastercard’s March 17 agreement to buy stablecoin infrastructure provider BVNK for up to $1.8 billion, including $300 million in contingent payments. The deal is expected to close in late 2026, subject to approvals.

Agentic AI is shifting commerce from human-initiated checkout toward software that can search, decide and pay within a user’s mandate. That raises new demands around identity, authorization, fraud controls and disputes, making payment infrastructure a strategic battleground. Mastercard is extending its card network, tokenization and transaction data into the “prompt economy,” seeking to preserve its role as a trusted intermediary as banks, Visa and technology platforms build competing rails for autonomous commerce.

Mastercard launched Agent Pay for Machines on June 10, 2026, enabling continuous machine-to-machine payments across cards, bank accounts and stablecoins, including microtransactions worth fractions of one cent. More than 30 early participants include Adyen, Stripe, Coinbase and Cloudflare. On the company’s July 30 earnings call, Chief Executive Michael Miebach said Mastercard’s global network and data scale should keep it central to agentic commerce despite competition from banks and technology giants, as the company moves the product from pilots toward broader adoption.

Agentic AI payments allow AI systems, with consumers’ authorization, to search for options, make selections and complete payments on their behalf. The key challenges are authenticating the agent’s identity, protecting card data and preserving user control. Mastercard launched Agent Pay in the United States in 2025 and expanded it to Asia-Pacific in 2026. Its collaboration with DBS Bank (Taiwan) marks the formal entry of AI-powered commerce into real-world transactions in Taiwan.

On March 26, 2026, Mastercard and DBS Bank (Taiwan) announced that they had completed Taiwan’s first authenticated agentic AI transaction. The transaction amount was not disclosed. A consumer entered the pickup and drop-off locations, time and number of passengers for an airport transfer into an AI chat interface. After matching available options, the AI booked and paid for the consumer’s selection with authorization using a linked Mastercard, while a Mastercard Agentic Token authenticated the agent’s identity.

AI agent payments allow systems not only to recommend products but also to complete transactions autonomously within spending limits and authorizations set by users. The significance of the collaboration between Banco Santander and Mastercard lies in bringing this form of agentic commerce into a regulated banking framework and testing whether existing payment networks can maintain security, privacy and consumer protection.

On March 2, 2026, Banco Santander used Mastercard Agent Pay in a controlled environment, with PayOS providing end-to-end orchestration, to complete Europe’s first live transaction initiated and executed by an AI agent through Santander’s existing payment infrastructure. The companies did not disclose the transaction value and stressed that the project remains a pilot and has not been commercially launched. They plan to expand testing and explore additional use cases.