a2aproject/A2A

What it solves

Agent2Agent (A2A) provides a standardized way for AI agents built on different frameworks (such as Google ADK, LangGraph, or BeeAI) and hosted on separate servers to communicate and collaborate. It prevents agents from becoming isolated silos by allowing them to work together as peers rather than just as tools, all while protecting their internal memory, proprietary logic, and tool implementations.

How it works

A2A establishes a common language for agents using JSON-RPC 2.0 over HTTP(S). It uses "Agent Cards" to allow agents to discover each other's capabilities and connection information. The protocol supports multiple interaction modes, including synchronous request/response, streaming via Server-Sent Events (SSE), and asynchronous push notifications, enabling the exchange of text, files, and structured JSON data.

Who it’s for

Developers building agentic applications who need their agents to interoperate with agents from other companies or ecosystems, as well as those creating multi-agent systems where specialized agents must collaborate on complex tasks.

Highlights

• Cross-Framework Interoperability: Supports agents regardless of the underlying framework used to build them.
• Agent Discovery: Uses standardized Agent Cards to detail capabilities.
• Privacy-Preserving Collaboration: Agents can collaborate without exposing internal state or memory.
• Enterprise-Ready: Built with a focus on security, authentication, and observability.
• Multi-Language SDKs: Provides official SDKs for Python, Go, JS, TypeScript, .NET, Java, and Rust.