Agent2Agent Protocol (A2A)

Canonical version: Agent2Agent Protocol (A2A).

Agent2Agent (A2A) is an open protocol for agent-to-agent interoperability, introduced by Google in 2025 and later donated to the Linux Foundation. It lets independent AI Agents, built by different teams on different frameworks, discover each other, exchange messages, and delegate tasks over a standard interface. Each agent publishes an Agent Card describing its identity and capabilities; others use it to decide what to hand off.

How it fits with MCP

A2A and the Model Context Protocol (MCP) are complementary, not competing:

• MCP connects an agent to tools, data, and context (vertical: agent to capability)
• A2A connects agents to each other (horizontal: agent to agent)
• Discovery layers like the Agentic Resource Discovery Specification (ARD) sit above both, helping a client find which agent or tool to use in the first place

It appears in agent platforms such as the MCP Gateway Registry (agent registry over A2A) and Strands Agents (multi-agent orchestration).

Related

• AI Agents
• Agentic Engineering
• Model Context Protocol (MCP)
• Agentic Resource Discovery Specification (ARD)
• MCP Gateway Registry
• Strands Agents
• Large Language Models (LLMs)

About Sébastien

Ready to get to the next level?