Mastercard has launched Agent Pay for Machines, a protocol designed for a future of continuous, always-on payments between autonomous AI agents. The system targets fast, high-frequency agent-to-agent transactions and micropayments, extending Mastercard's broader Agent Pay framework into machine-driven commerce.
Built for agent-to-agent commerce
As AI agents take on more tasks that involve buying goods, services and compute, payment networks are racing to provide trusted rails for those interactions. Agent Pay for Machines is aimed at scenarios where software agents transact with one another at high speed, requiring infrastructure that can handle frequent, low-value transfers.
Industry support
Mastercard listed a wide group of initial participants and supporters, spanning payments, crypto and infrastructure firms:
- Payments and processing: Adyen, Checkout.com and Stripe.
- Crypto and digital assets: Aave Labs, Anchorage Digital and Coinbase.
- Infrastructure: Alchemy, Cloudflare and Ant International.
The competitive landscape
The launch lands amid intensifying competition over agentic commerce. Visa has advanced its own Intelligent Commerce approach and announced a direct partnership with OpenAI to embed payment capabilities into its experiences. Stripe, meanwhile, has introduced Shared Payment Tokens that let agents initiate payments with a customer's permission without exposing underlying credentials, and has expanded support to network-led capabilities including Mastercard's framework.
How the pieces fit
Different players are emphasizing different design choices:
- Mastercard binds tokenized credentials to specific agents, merchants and consent policies.
- Visa accepts payments across multiple protocols through a single integration.
- Stripe offers a payment primitive that abstracts away sensitive credentials.
Market potential
Analysts see a large opportunity. McKinsey has projected that AI agents could account for $1 trillion in US-based transactions by 2030, while more conservative estimates from Morgan Stanley and Bain place agentic activity near 15 to 17% of US e-commerce. Those figures help explain why the major networks are moving quickly to define the standards.
For Mastercard, Agent Pay for Machines represents a bet that agent-to-agent payments will become a meaningful share of commerce, and that the firms providing secure, policy-aware rails will be positioned at the center of it. The protocol's adoption will hinge on whether developers and merchants embrace it as agentic commerce moves from concept to deployment.
Trust and control in agent payments
A central challenge in agentic commerce is ensuring that autonomous software acts within the boundaries a user or business intends. Payment networks are addressing this by binding credentials to specific agents and attaching consent policies that define what an agent is permitted to spend, with whom and under what conditions. These guardrails are meant to prevent runaway or unauthorized transactions while still allowing agents to operate quickly and independently.
The competition among Mastercard, Visa and Stripe is in part a race to establish the standards that merchants and developers will build around. With analyst projections pointing to large potential transaction volumes from AI agents in the coming years, the firm or protocol that becomes a default for secure agent payments could gain a durable advantage. As deployments expand, the practical reliability and security of these systems, rather than their announcements, will determine which approaches gain lasting traction.