A2AMCPDemo

# A2A MCP Demo A multi-agent system demo that uses the [A2A protocol](https://a2aproject.github.io/A2A/latest/) and an A2A [MCP](https://modelcontextprotocol.io/introduction) server. ## What's Included - **MCP Server** (`mcp_server.py`): A FastMCP server that provides tools for agent discovery, communication, and session management via A2A. - **Orchestrator Agent** (`orchestrator_agent.py`): A [LangGraph](https://langchain-ai.github.io/langgraph/) agent uses [A2ANet](https://github.com/A2ANet/A2ANetPython) to easily implement A2A. Coordinates multiple AI agents using MCP tools. - **Tavily Agent** (`tavily_agent.py`): A [LangGraph](https://langchain-ai.github.io/langgraph/) agent uses [A2ANet](https://github.com/A2ANet/A2ANetPython) to easily implement A2A. Web search agent. ## Quick Start 1. Clone the repository: `git clone https://github.com/A2ANet/A2AMCPDemo.git` 2. Create an `.env` file from `.env.example` 3. Set `ANTHROPIC_API_KEY` and `TAVILY_API_KEY` 4. Optionally, set `LANGSMITH_*` for LangSmith tracing 5. Install [uv](https://docs.astral.sh/uv/) 6. Run the Orchestrator agent: `uv run --env-file .env orchestrator_agent.py` 7. Run the Tavily agent: `uv run --env-file .env tavily_agent.py` That's it! The Orchestrator agent will be running on `http://localhost:8000` and the Tavily agent on `http://localhost:8001`. Test the Orchestrator Agent with the [Agent2Agent (A2A) UI](https://github.com/A2ANet/A2AUI) or [A2A Protocol Inspector](https://github.com/a2aproject/a2a-inspector). Ask it to "discover an agent at http://localhost:8001", then ask it "what's the weather in London?". It will use the A2A MCP server to fetch the Tavily Agent's agent card, then to send it a message. The Tavily Agent will then use the [Tavily Search API](https://www.tavily.com/) to search for the weather in London, and send an `Artifact` back for the Orchestrator agent to display. ## License Distributed under the terms of the [Apache-2.0](https://spdx.org/licenses/Apache-2.0.html) license.