ai-boost/awesome-a2a

What it is: A hand-picked, well-organized list of links and resources about the Agent2Agent (A2A) Protocol — an open standard that lets different AI agents talk to each other securely and work on tasks together, even if they're built by different vendors or on different frameworks.

This repository doesn't contain agent code itself. It's a guide: someone new to A2A can come here to learn what the protocol is, read the official documentation, find sample code, discover community libraries, and track key announcements. It's formatted as an "awesome list," the popular community style for curated resource collections.

What A2A is (in plain terms): Today, AI agents (like assistants or automation bots) often live in silos — one vendor's agent can't easily call another's. A2A defines common rules of communication so agents from different systems can discover each other, exchange requests, and hand off tasks. It emphasizes being simple, using existing web standards; being enterprise-ready (security, auth, privacy); handling long-running tasks and human-in-the-loop workflows; supporting many content types (text, files, forms); and keeping agent internals opaque — they share results, not their internal logic.

How it works at a high level: Agents publish an Agent Card (a JSON file describing what they can do and how to reach them). A client agent sends a Task request (with messages broken into Parts) to a remote agent server over HTTP/JSON-RPC. The server processes it, updates the task status, and returns results (called Artifacts). For long tasks, the server can stream updates or send push notifications.

What's inside the list:

• Introductory material — explanations of key concepts, plus links to official documentation.
• Getting-started path — a step-by-step suggested route for newcomers, from reading the basics to running the official sample agents.
• Official resources — links to the protocol website, GitHub repo, the v1.0.0/v1.0.1 release notes, blog posts (including the March 2026 v1.0 release announcement and the Linux Foundation anniversary press note), and news about adoptions (e.g., Microsoft-SAP integration, cloud platform support).
• Specification & core concepts — links to technical docs on Agent Cards, discovery, push notifications, enterprise readiness, the official extension mechanism, and experimental extensions (like OID4VP auth and a custom low-latency transport binding).
• Implementations & libraries — a large table of official sample agents across TypeScript, Java, .NET, and Python (e.g., movie-recommendation agent, LangGraph agent, Semantic Kernel demo), plus community SDKs, frameworks, and platform integrations.
• Tools & utilities, tutorials & articles, demos & examples, and a section on related protocols & concepts (like MCP).

Who it's for: Developers who want to build A2A-compatible agents, researchers tracking the protocol's evolution, or anyone curious about how AI agents can interoperate. The list is maintained by the community under the MIT license.

Bottom line: If you want to know everything about how AI agents can talk to each other via the A2A protocol — from theory to working code — this is the starting point. It's a curated map of the ecosystem, not the software itself.