A curated list of SDKs, frameworks, agents, bridges, tools, and resources for the Agent-to-Agent (A2A) protocol — the open standard for agent interoperability introduced by Google in 2025.
Each A2A agent exposes a /.well-known/agent.json Agent Card and communicates over HTTP/SSE/JSON-RPC. MCP handles tool context; A2A handles agent-to-agent task routing. They are complementary, not competing.
Browse the full commercial catalog → a2abay.com
- Spec & Docs
- Official SDKs
- Community SDKs & Libraries
- Frameworks (A2A-ready)
- MCP ↔ A2A Bridges
- Tools & Dev Utilities
- Infrastructure & Discovery
- Sample Agents (Official)
- Community Agents
- Learning Resources
- A2A Protocol Specification — Canonical spec repo. Agent Card schema, task lifecycle, transport definitions, JSON-RPC message format.
- A2A Protocol Docs — Official documentation hub with guides and implementation details.
- A2A Samples Repo — All official reference agents across Python, TypeScript, Java, Go, and .NET.
| Language | Repo | Notes |
|---|---|---|
| Python | a2a-python | Streaming, Agent Cards, push notifications, framework integrations |
| TypeScript / JS | a2a-js | Express + Node.js, web apps and toolchains |
| Go | a2a-go | Lightweight concurrent A2A services |
| Java | a2a-java | JVM, typed clients, Quarkus reference |
| .NET / C# | a2a-dotnet | ASP.NET Core, typed task flows, NuGet packages |
- python-a2a
Python979★— The most complete community library: workflow engine, AI router, Agent Flow UI drag-and-drop, native LangChain, MCP v2.0 with GitHub/Browserbase/Filesystem providers, full CLI. - Legion A2A
Python— Strongly typed with Pydantic v2, dataclasses, and robust serialization support. - a2a_min
Python— Minimal implementation covering client/server basics, task lifecycle, streaming, and push updates. Good for learning the protocol from scratch. - a2a-net
C#— Modular: Core, Client, Server, HTTP + gRPC + JSON-RPC, Redis persistence, Quartz scheduling. Assemble only what you need. - a2a4j
Java— Spring Boot starters, BOM, Reactor SSE streaming, A2A 1.0 spec-faithful, Maven Central. - a2ajava
Java— Spring Boot annotations, WebSockets, and MCP integration. - Agentic
Rust— MCP client/server crate in active development, ergonomic building blocks, A2A on roadmap.
| Framework | Support | Language | Notes |
|---|---|---|---|
| Google ADK | native |
Python | First-class A2A, reference for official samples |
| LangGraph | native |
Python | A2A task routing since v0.3 |
| CrewAI | native |
Python | Multi-agent orchestration with A2A |
| AutoGen / AG2 | native |
Python | Expose any AutoGen agent as A2A server in 3 lines |
| Semantic Kernel | native |
Python / C# | Documented by Microsoft, travel manager sample |
| BeeAI Framework | native |
Python | ACP-to-A2A evolution, LRU session memory, Ollama |
| Genkit | native |
TypeScript | Official Google samples (movie, coder, editor) |
| Quarkus LangChain4j | native |
Java | Cross-language pipeline, swappable agents |
| Azure AI Foundry | plugin |
Python | Playwright + Azure Functions SSE MCP + Gradio |
| Cognisphere | community |
Python | Cognitive arch on ADK: 5 specialized agents, persistent embeddings, AIRA network |
Use a bridge when you want to expose an MCP server as an A2A agent, or call A2A agents from Claude Desktop, Cursor, or Zed.
- A2A MCP Server Bridge — Most complete on PyPI. 5 tools,
pip install, Smithery-compatible, multi-transport (stdio + streamable-HTTP). Works with Claude, Cursor, Windsurf. - Python MCP_A2A Bridge — Lightweight 3-tool bridge (register, list, call). Hackable, not production-grade. Good starting point.
- A2A MCP Without Framework — Low-level sample showing A2A + MCP without any framework. Best for understanding the raw protocol layer.
- Python A2A MCP Sample — ADK + LangGraph + MCP complementing each other in one codebase.
- A2A Inspector
official— Validation UI for Agent Cards, task runs, streaming events, spec compliance, and raw protocol traffic. - A2A Integration Test Kit
official— Compatibility kit for verifying cross-SDK A2A interoperability across transports and versions. - Cisco A2A Scanner — Open-source security scanner for auditing A2A endpoints, Agent Cards, and registry-facing trust surfaces.
- A2AServer — Full-stack framework: Python backend + Node.js frontend, multi-agent coordination, MCP tools via JSON config, RAG and DeepSearch agents ready to deploy.
- Swival — Open-source CLI coding agent with A2A and MCP support across many model providers.
- AGNTCY — Open infrastructure stack for agent discovery, identity, messaging, and observability across any vendor or framework.
- A2A Registry — Global public registry. Register in under 5 minutes, machine-facing discovery via
POST /a2a/discover, routing without hardcoded URLs. - A2A.ac — Hosted directory of public A2A agents from official and community sources. Browsable with zero setup.
- awesome-a2a — Curated directory of deployable A2A-compliant servers classified by business category, language, and scope.
- Pinchwork — Agent-to-agent task marketplace: register in one curl, post tasks, pick up work, earn credits. Go CLI + Docker self-hosting.
All from a2aproject/a2a-samples unless noted.
- HelloWorld Agent — Under 50 lines. The absolute starting point for any A2A server.
Custom - LangGraph Currency Agent — Multi-turn, SSE streaming, checkpoint memory, push notifications, JWK auth. Most feature-complete Python sample.
LangGraph - ADK Currency Agent — FastMCP + Frankfurter API + ADK + A2A. Reference for the MCP+ADK+A2A stack.
Google ADK - Expense Reimbursement Agent — DataPart pattern, dynamic forms, multi-turn information gathering, approval logic.
Google ADK - Purchasing Concierge — ADK root agent orchestrating Pizza Seller (LangGraph) and Burger Seller (CrewAI) over A2A. Cross-framework coordination demo.
Google ADK - Pizza Seller Agent — Answers menu queries and accepts orders directly. No confirmation step.
LangGraph - Burger Seller Agent — Requests user confirmation before finalizing orders, preserving agent intent through A2A.
CrewAI - CrewAI Image Generation Agent — Prompt-to-image returning image artifacts. Reference for non-textual artifact output.
CrewAI - Semantic Kernel Travel Manager — Routes between CurrencyExchangeAgent and ActivityPlannerAgent, multi-turn, ephemeral chat history.
Semantic Kernel - Azure AI Foundry Sample — Playwright agent + Azure Functions SSE MCP server + Gradio interface.
Azure AI Foundry + SK - BeeAI Chat Sample — 4 built-in tools (search, weather, Wikipedia, ThinkTool), Ollama, LRU session memory.
BeeAI - GitHub Agent Sample — GitHub-centric automation and repo tasks.
OpenAI Function Calling - Travel Planner Agent — Multi-step travel planning and itinerary assistance.
OpenAI-compatible - Airbnb Planner Multi-Agent — Accommodation planning with cross-agent coordination.
Google ADK
- JS Movie Agent — TMDB API, movie search, cast lookups. Clearest beginner TypeScript sample.
Genkit - JS Coder Agent — Returns generated code as A2A artifacts. Reference for the artifact pattern.
Genkit - JS Content Editor — Proofreads copy in the cross-language content creation pipeline. Swappable with the Java editor.
Genkit
- Java Weather MCP Sample — Weather data + MCP tooling + A2A orchestration.
Quarkus LangChain4j - Java Content Writer — Generates long-form copy from a structured outline.
Quarkus LangChain4j - Java Content Editor — Directly swappable with the TypeScript editor via A2A.
Quarkus LangChain4j
- .NET Semantic Kernel Demo — A2A + Semantic Kernel in C#. Part of the a2a-dotnet SDK.
Semantic Kernel
- AG2 Mypy Tool Agent — Multi-agent coder+reviewer workflow, MCP-backed mypy validation, streaming status updates.
AutoGen
- Cerebrus Pulse — Live crypto intelligence on 50+ Hyperliquid perps via A2A, ACP Virtuals, and x402 USDC. 12 endpoints, pay-per-query from $0.008 USDC.
hostedpaid
- DeepLearning.AI A2A Course — Short course built with Google Cloud and IBM Research. Free.
- Google Cloud: Develop an A2A Agent — Official tutorial for building A2A-compatible agents on Google Cloud.
- Microsoft Foundry: Add an A2A Endpoint — Official guide for exposing Foundry Agent services over A2A.
- IBM: A2A Business Workflow Tutorial — Build and deploy A2A agents for complex business workflows.
- KDnuggets: A2A vs MCP Explained — Solid conceptual overview of how A2A and MCP complement each other.
- Camunda: MCP, ACP, and A2A — Comparison of the three main inter-agent communication standards.
- A2ABay Blog — Protocol notes, marketplace analysis, and practical guidance for buyers and builders.
PRs welcome. Format:
- [Name](url) `Language` — Short factual description.
Criteria: repo must be public, have a README in English, and relate directly to the A2A protocol or its ecosystem. No promotional language.
See CONTRIBUTING.md for details.
Commercial catalog and marketplace → a2abay.com