A2A Protocol
also: agent-to-agent protocol · Agent2Agent
The open standard for agents talking to other agents — discovery, task delegation, and result exchange.
A2A (Agent2Agent) standardizes how independent AI agents discover each other and collaborate: an agent publishes an agent card describing its capabilities, receives tasks over a common HTTP/JSON-RPC interface, streams progress, and returns artifacts. Where MCP connects one agent to its tools and data, A2A connects peer agents across vendors and frameworks — the two protocols are complementary layers of the emerging agent stack.
Worked example: agent-to-agent communication — protocols and message formats that let independent AI agents discover each other, delegate tasks, and exchange results, so a system can be composed of specialized agents rather than one monolith. Gotcha: multi-agent is not automatically better — every hand-off adds latency, cost, and a place for miscommunication, and errors compound across agents; use it when tasks genuinely decompose into independent specialists, not to split a job one well-equipped agent could do more reliably.