MCP Server - LangChain

LangChain 1.4 and later supports MCP servers through its first-party MCPAdapter. Install LangChain with the MCP extra and the Anthropic integration:


pip install "langchain[mcp]>=1.4.0" langchain-anthropic

Create an authenticated FastMCP client, discover the Xweather tools, and pass them to a LangChain agent:


import asyncio
import os
 
from fastmcp.client import Client
from langchain.agents import create_agent
from langchain.mcp import MCPAdapter
from langchain_anthropic import ChatAnthropic
 
 
async def main():
    url = (
        "https://mcp.api.xweather.com/mcp"
        "?include_tools=xweather_get_current_weather,"
        "xweather_get_forecast_weather,"
        "xweather_get_current_weather_impacts"
    )
    client = Client(
        url, auth=os.environ["XWEATHER_API_KEY"]
    )
 
    async with MCPAdapter(client) as adapter:
        tools = await adapter.list_tools()
        agent = create_agent(
            model=ChatAnthropic(
                model="claude-sonnet-5", api_key=os.environ["ANTHROPIC_API_KEY"]
            ),
            tools=tools,
            system_prompt=(
                "You are a weather analyst. Use Xweather tools when needed "
                "and identify the time period and location in your answer."
            ),
        )
        response = await agent.ainvoke(
            {
                "messages": [
                    {
                        "role": "user",
                        "content": "Draft an operations brief for ORD tonight.",
                    }
                ]
            }
        )
 
    print(response["messages"][-1].content)
 
 
asyncio.run(main())

Tips

Keep discovery and agent execution inside the MCPAdapter context so the connection remains open.
Restrict exposed tools with Xweather’s include_tools or include_tags query parameters.
The langchain.mcp namespace is currently beta and may change in later LangChain releases.