Google proposed Agent2Agent as a shared interface for agent discovery, messaging, and task delegation. Implementations could coordinate across framework boundaries through a common task interface.
Agent2Agent Protocol (A2A)
Also known as: A2A · Agent-to-Agent Protocol
Plain English
Agent2Agent is an open protocol for software agents to discover one another, request work, and receive results. The communicating systems can use different models or internal frameworks.
Technical Overview
An agent card advertises capabilities and describes how a client can contact the service. The protocol defines messages, tasks and their state, and the artifacts returned by an agent. This complements the Model Context Protocol, which standardizes how applications access tools and resources.
Timeline (4)
2025
Milestone- Ecosystem
The Linux Foundation launched the A2A project after Google contributed the protocol. This established a shared venue for maintaining the specification and related software.
2026
ReleaseThe project published A2A v1.0 as its first stable specification. It defines shared task semantics across multiple transport bindings and a way to negotiate protocol versions.
- Ecosystem
The Agentic AI Foundation announced that A2A was joining as a hosted project, bringing agent-to-agent messaging and the MCP tool protocol under the same foundation.