Agent-to-Agent Protocol (A2A)
protocolAssess
Agent-to-Agent (A2A) is an open, HTTP-based protocol that enables AI agents—built on different frameworks to coordinate tasks, exchange information, and collaborate. Backed by companies like Google, Salesforce, and LangChain, A2A standardizes inter-agent communication and discovery, preparing the way for modular, scalable multi-agent systems. This protocol is especially suitable when agents must operate independently but collaborate via clearly defined interfaces.
While MCP standardizes LLM-to-environment interaction, A2A focuses on agent-to-agent coordination. These protocols are complementary: MCP enables agents to reason with tools and context, while A2A enables those agents to talk to each other.
Key Concepts
- A2A introduces three core actors: User (initiates requests), A2A Client (that makes requests on the user's behalf) and A2A Server (remote agent that fulfills requests and returns results)
- AgentCard: JSON descriptor of an agent’s identity, capabilities, skills, endpoint, and auth requirements.
- Discovery: Agents can be found via
.well-known
URLs, registries, or static configuration. - Communication: JSON-RPC over HTTP(S) using tasks and artifacts. Supports:
- Synchronous (
sendTask
) - Asynchronous polling
- Streaming (SSE)
- Push (webhooks)
- Synchronous (
- Security: AgentCard-defined auth schemes (API keys, OAuth2)