A2A Adapter

Open Source A2A Protocol Adapter SDK — Convert any AI agent framework into a fully compliant A2A Protocol server in 3 lines of code.

Installation


pip install a2a-adapter

Quick Example


from a2a_adapter import N8nAdapter, serve_agent
 
adapter = N8nAdapter(webhook_url="http://localhost:5678/webhook/agent")
serve_agent(adapter, port=9000)

That’s it. Your n8n workflow is now an A2A-compliant server with:

Auto-generated Agent Card at /.well-known/agent.json
Task management and lifecycle
Push notifications
A2A JSON-RPC endpoint

See the Quick Start to get your first agent running.

Framework Support

Framework	Adapter	Streaming	Status
n8n	`N8nAdapter`	No	Stable
LangChain	`LangChainAdapter`	Auto-detected	Stable
LangGraph	`LangGraphAdapter`	Auto-detected	Stable
CrewAI	`CrewAIAdapter`	No	Stable
OpenClaw (openclaw.ai — AI coding agent)	`OpenClawAdapter`	No	Stable
Ollama	`OllamaAdapter`	Yes	Stable
Custom Function	`CallableAdapter`	Optional	Stable
Custom Class	`BaseA2AAdapter`	Optional	Stable

Key Benefits

3-line setup — Expose any agent as A2A-compliant with minimal code
Built-in adapters — n8n, LangChain, LangGraph, CrewAI, OpenClaw, Ollama, and callable functions
Streaming support — auto-detected for LangChain and LangGraph, always on for Ollama, optional for callables
ASGI deployment — use to_a2a() when you want to run under your own server
Extensible — create custom adapters or register third-party ones

Three Levels of Control

Level	Method	Use Case
Level 1	`invoke(input) -> str`	90% of use cases — just return text
Level 2	`stream(input) -> AsyncIterator`	Streaming frameworks (LangChain, Ollama)
Level 3	Implement `AgentExecutor` directly	Full SDK access for advanced scenarios

Architecture


Layer 1: Public API
    serve_agent() / to_a2a() / build_agent_card()
         |
Layer 2: A2A SDK
    DefaultRequestHandler, TaskStore
         |
Layer 3: Bridge
    AdapterAgentExecutor (thin ~80-line bridge)
         |
Layer 4: Adapter Interface
    BaseA2AAdapter (invoke / stream / cancel)
         |
Layer 5: Framework Drivers
    N8nAdapter, LangChainAdapter, CrewAIAdapter, ...

The key design insight is the bridge layer (AdapterAgentExecutor) that translates between the simple adapter interface (just implement invoke()) and the full A2A SDK event-driven model. This means adapters stay simple while getting full protocol compliance for free.