Stripe deploys 6 new protocols to enable 250M users to authorize AI agents for real-time token payments

The fundamental architecture of Internet commerce is undergoing a structural shift as artificial intelligence agents transition from passive tools to active economic participants capable of placing orders, deploying applications, and consuming tokens. This evolution necessitates a redefinition of economic infrastructure, moving beyond the traditional model where payment systems served only human users and software companies to a new paradigm supporting machine entities. Agents now require the ability to interpret pricing, execute payments, access wallets, and deploy services autonomously, compelling AI enterprises to establish novel billing, risk control, and settlement mechanisms centered on token consumption. Data compiled by Woofun AI indicates that initiatives such as the Machine Payments Protocol, Link, Stripe Projects, Metronome, Tempo, and streaming payments collectively signal a trend where agents act as both buyers and builders, creating software while simultaneously consuming resources. Consequently, the legacy SaaS business model reliant on seat-based fees, post-facto invoicing, and manual reconciliation is being displaced by real-time metering, instant settlement, and machine-readable frameworks.

Over the past few years, the focus has been on building economic infrastructure to assist the world's fastest-growing AI companies through payment, billing, checkout, fraud prevention, and tax systems, while also deploying AI within payment processes to boost profitability.

However, the definition of AI economic infrastructure has evolved significantly in the last six months as agents emerge as distinct actors on the Internet.

This shift introduces two critical objectives: enabling agents to purchase and build on behalf of individuals and businesses, and assisting companies in adapting to an economic model driven by token consumption. The goal is to empower agents as independent actors while helping businesses monetize products created and consumed by these autonomous entities. E-commerce was originally designed for human interaction involving browsing, clicking pricing pages, and entering credit card details, but smart agents operate through a programmatic approach that requires understanding service needs, fees, and payment execution without human intervention.

To address the need for programmatic payment acceptance, a partnership with Tempo resulted in the creation of the Machine Payments Protocol. This protocol allows businesses to accept payments directly from smart agents without human intervention, eliminating account registrations, checkout pages, and human-in-the-loop requirements in favor of a machine-readable payment method. Once businesses can accept payments from agents, the critical challenge becomes how consumers can securely authorize agents to spend on their behalf. This is where the consumer wallet Link becomes pivotal. Currently, over 250 million people utilize Link, and its capabilities are being extended to serve smart agents, allowing users to securely authorize agents to make payments while retaining ultimate control. Woofun AI notes that this extension ensures humans maintain oversight while granting agents the necessary spending authority.

Empowering agents extends beyond purchasing to their ability to build and deploy applications. While vibe-coding has become accessible, vibe-deploying remains a bottleneck requiring significant manual effort for account creation, service configuration, credential management, and API access. Stripe Projects addresses this by enabling developers and their smart agents to register, manage, and integrate necessary services directly through the command line, aiming to make deployment as seamless as coding. As agents assume roles as buyers and builders, businesses must adapt to a dynamic cost environment where product reasoning and token costs fluctuate. In the traditional SaaS model, serving additional users incurs minimal marginal cost, but in AI products, every prompt, API call, and agent task carries a real marginal cost, rendering seat-based billing insufficient.

Consumption-based billing has emerged as the key to AI monetization, allowing companies to charge based on usage, workflows, outcomes, or other units that reflect true value. Lovable exemplifies this transition, initially adopting a simple subscription model before evolving to charge based on consumed AI tokens once usage limits were exceeded. ElevenLabs followed a similar trajectory, shifting from subscriptions to usage-based pricing as product and customer patterns matured.

However, the viability of consumption-based billing depends on the ability to recoup funds, as fraudsters increasingly target tokens rather than just funds or credentials. If bad actors can register accounts, rapidly consume tokens, and vanish before the billing cycle ends, the economic model collapses. Woofun AI analysis suggests that token theft represents one of the most under-discussed yet critical risks in the current AI industry landscape.

Stripe Radar addresses these threats by assessing new accounts in real-time, predicting abuse of free trials, and identifying unpaid usage risks as consumption accumulates. The complexity intensifies when the customer is an intelligent agent capable of consuming tokens at machine speed, making traditional billing methods unsustainable. Companies face a dilemma between requiring upfront payments, which protects revenue but degrades customer experience, or allowing usage to accumulate, which improves experience but risks non-payment. The optimal solution involves real-time usage tracking and real-time payments, a capability achieved through the combination of Metronome and Tempo. Metronome, acquired by Stripe, handles complex usage-based billing models with real-time tracking, while Tempo supports low-cost, high-frequency stablecoin payments and instant settlement.

This integration enables AI companies to charge in real-time as tokens are consumed, eliminating the trade-off between hard caps and uncollectible invoices. This new business model, termed Streaming Payments, is tailored for AI-native businesses adapting to machine-speed software consumption. The broader implication is that AI economic infrastructure must now support a commercial system for agents, providing wallets, deployment capabilities, token-based billing, fraud prevention, and streaming payment features. As AI reshapes business operations and enterprise structures, the underlying infrastructure must evolve accordingly to support this new reality where agents are central to the economic ecosystem.