How do I deploy an MCP server on LangSmith Deployments with my agent?

While this setup is not recommended for production, it’s possible to run an MCP server within a LangSmith Deployment for proof-of-concept (PoC) or demo purposes.

This example is adapted from the LangChain MCP Adapters README.

Recommended approach is to stand up a separate MCP server elsewhere and deploy it independent of a LangGraph Deployment

Example: Running an MCP Server with a LangGraph Agent

If you want to run a LangGraph agent that uses MCP tools in a LangGraph API server, you can use the following setup:

# graph.pyfrom contextlib import asynccontextmanagerfrom langchain_mcp_adapters.client import MultiServerMCPClientfrom langgraph.prebuilt import create_react_agent

async def make_graph():
    client = MultiServerMCPClient(
        {
            "weather": {
                # make sure you start your weather server on port 8000
                "url": "http://localhost:8000/mcp",
                "transport": "streamable_http",
            },
            # ATTENTION: MCP's stdio transport was designed primarily to support applications running on a user's machine.
            # Before using stdio in a web server context, evaluate whether there's a more appropriate solution.
            # For example, do you actually need MCP? or can you get away with a simple `@tool`?
            "math": {
                "command": "python",
                # Make sure to update to the full absolute path to your math_server.py file
                "args": ["/path/to/math_server.py"],
                "transport": "stdio",
            },
        }
    )
    tools = await client.get_tools()
    agent = create_react_agent("openai:gpt-4.1", tools)
    return agent

In your langgraph.json make sure to specify make_graph as your graph entrypoint:

{
  "dependencies": ["."],
  "graphs": {
    "agent": "./graph.py:make_graph"
  }
}

The second MCP server above shows how to point to a fastmcp application in a file in your repo and have it run everytime.

There are a number of reasons why not to do this in production!! Highlighted in this thread: https://langchain.slack.com/archives/C089RDWTKU4/p1761774370247319