Google announced the Agent2Agent protocol, or A2A, on April 9, 2025 at Google Cloud Next. Where the Model Context Protocol standardizes how a single agent connects to tools and data, A2A standardizes how separate agents - potentially built by different vendors on different frameworks - talk to one another, exchange information securely, and hand off tasks. Google framed the two as complementary: MCP gives an agent tools and context, A2A lets agents collaborate.
The design rests on a few choices. Agents advertise what they can do through “Agent Cards,” JSON documents that let a client agent discover and pick the right collaborator. The protocol builds on existing web standards - HTTP, Server-Sent Events, and JSON-RPC - rather than inventing new transport, and it is designed for enterprise-grade authentication, long-running tasks with real-time progress, and multiple modalities including text, audio, and video. More than fifty technology partners and a long list of consulting firms signed on at launch.
A2A’s significance is that it treats interoperability between agents as infrastructure. In June 2025 Google contributed the protocol to the Linux Foundation under an Apache 2.0 license, and adoption spread across cloud platforms and vendors, making it one of the early standards for the emerging world of many specialized agents working together rather than one monolithic assistant.