The emerging infrastructure for AI-driven commerce is forming a layered architecture rather than a single standard. Key protocols include Anthropic’s Model Context Protocol (MCP) for data access and Google’s Agent2Agent (A2A) for communication between agents.
Commerce-specific standards are also expanding. OpenAI and Stripe introduced the Agentic Commerce Protocol (ACP) for discovery and checkout, while Google and Shopify developed the Universal Commerce Protocol (UCP) to manage carts, identity, and loyalty programs.
Security and authorization protocols are evolving to handle machine-mediated transactions. Google’s Agent Payments Protocol (AP2) uses cryptographically signed mandates to verify consumer intent, while Visa’s Trusted Agent Protocol (TAP) and Mastercard’s Agent Pay focus on identity verification.
Stripe and Tempo have launched the Machine Payments Protocol (MPP) to facilitate machine-to-machine transactions. These developments address the need to distinguish legitimate AI agents from malicious bots, a growing concern as AI-driven fraud becomes more persistent.
A PYMNTS Intelligence report indicates that 56% of consumers are willing to let agents compare products, and 35% would grant them access to payment credentials. However, a readiness gap persists, with 75% of tech firms familiar with agentic AI compared to 33% of goods firms.
