Agent-to-Agent Protocol (A2A)
A standard that lets AI agents built by different companies communicate and delegate tasks to each other.
What is Agent-to-Agent Protocol (A2A)?
A2A is a standard that lets AI agents built by different companies communicate with each other and hand off tasks, even if the agents were built on completely different software frameworks. For example, a travel-planning agent from one company could ask a separate hotel-booking agent from another company to check availability, and A2A defines the common "language" both agents use to make that request and understand the response. Without a standard like this, agents from different vendors generally can't talk to one another at all.
Why it matters
Google introduced A2A on April 9, 2025, and donated the protocol to the Linux Foundation on June 23, 2025, at the Open Source Summit North America, establishing a vendor-neutral Agent2Agent project. A2A runs over HTTP with JSON-RPC 2.0 and Server-Sent Events for streaming, and uses "Agent Cards" (JSON documents describing an agent's capabilities and authentication requirements) to let agents discover what other agents can do before delegating a task. It is released under an Apache 2.0 license, reached specification version 1.0 in 2026, and as of its one-year anniversary in April 2026 had grown to more than 150 supporting organizations, including Microsoft, AWS, Salesforce, SAP, ServiceNow, and IBM. Google's own Agent Payments Protocol (AP2) is built as an extension on top of A2A.