Hello everyone,
Google just introduced the Agent Payments Protocol (AP2), an open protocol designed to provide a secure and reliable framework for AI-driven commerce.
For those building payment applications using AI agents, a critical challenge is emerging: how do you securely handle payments initiated by a non-human agent?
Today’s payment systems assume a human is present and interacting with a trusted interface. This leaves critical gaps when an agent acts on a user’s behalf:
- How do you prove the agent had the user’s specific authority for a purchase?
- How do you protect against agent errors or “hallucinations” that could lead to incorrect purchases?
- How is accountability assigned in the case of fraud or error?
AP2 is designed to solve this problem by anchoring transactions to deterministic, non-repudiable proof of intent from the user, rather than relying on ambiguous, probabilistic AI outputs.
It works by defining a role-based architecture and using Verifiable Credentials (VCs)—cryptographically signed data objects—to manage the transaction flow. These VCs (like the Intent Mandate and Cart Mandate) create a clear, auditable trail that establishes authorization and accountability for all parties.
AP2 is available as an extension for the open-source A2A protocol and relies on MCP (Model Context Protocol).
And I invite you to explore the technical specification, documentation, and code samples now available on the public GitHub repository: https://github.com/google-agentic-commerce/AP2.
Happy building!