Stripe Builds for an AI-Powered Economy

Stripe is actively building for a future of "agentic commerce," where AI assistants autonomously purchase goods and services. This strategy involves creating the financial infrastructure to support a high volume of machine-to-machine transactions. Co-founder John Collison predicts a "torrent" of AI-driven commerce that moves beyond simple automation to independent financial decisions made by AI. To facilitate this, Stripe co-developed the Agentic Commerce Protocol (ACP) with OpenAI, creating a standard for AI agents to interact with and purchase from businesses. Early partners testing solutions based on this protocol include Microsoft Copilot, Anthropic, and Perplexity. This protocol aims to remove the friction of human-driven purchasing, potentially unlocking new economic activity. A key technical component is the Stripe Agent Toolkit, an SDK that allows developers to integrate Stripe's API with popular AI agent frameworks like LangChain and CrewAI using Python and TypeScript. This enables agents to perform tasks like creating payment links, managing product catalogs, or issuing virtual cards based on simple, plain-English commands. Stripe has also introduced a "Payments Foundation Model," an AI model trained on billions of transactions to combat fraud. This model dramatically improved fraud detection rates for large users from 59% to 97%. This specialized AI is crucial for securing the high-speed, high-volume transactions anticipated in the agent economy. To address the high cost of running AI models, Stripe launched a feature in preview that allows businesses to bill end-users for the large language model (LLM) tokens they consume, plus a markup. This turns a significant cost center into a potential revenue stream, a critical step for the financial viability of many AI startups. The company is also leveraging stablecoins like USDC as a primary settlement layer for their 24/7 availability and programmability. In collaboration with Paradigm, Stripe developed the Tempo blockchain, specifically designed to handle the high-frequency transactions expected from AI agents. To further support developers, Stripe offers tools like a VS Code extension with an AI Assistant for answering questions about the API and providing code suggestions. The company also provides a Model Context Protocol (MCP) that allows AI agents to interact with the Stripe API and search its knowledge base directly. While bullish on the long-term vision, Stripe's leadership acknowledges that the shift to fully autonomous agents will be gradual. Co-founders Patrick and John Collison have outlined five levels of agentic commerce, stating the industry is currently just on the edge of the first two, which involve more descriptive and situational purchasing prompts rather than full autonomy.