> ## Documentation Index
> Fetch the complete documentation index at: https://docs.langchain.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Model Context Protocol (MCP)

> Connect LangChain agents to MCP servers with the MCPAdapter, built on FastMCP.

[Model Context Protocol (MCP)](https://modelcontextprotocol.io) is an open protocol that standardizes how applications provide tools and context to language models. LangChain agents call tools defined on MCP servers through [`MCPAdapter`](https://reference.langchain.com/python/langchain/mcp/adapter/MCPAdapter), which discovers a server's tools and adapts them into LangChain tools you can pass straight to [`create_agent`](https://reference.langchain.com/python/langchain/agents/factory/create_agent).

[`MCPAdapter`](https://reference.langchain.com/python/langchain/mcp/adapter/MCPAdapter) is built on [FastMCP](https://gofastmcp.com), which handles transport inference, protocol negotiation, connection management, and authentication. This section covers the LangChain-specific layer and links out to the FastMCP client documentation for the connection details underneath.

<Note>
  The `langchain.mcp` namespace requires `langchain[mcp]>=1.4.0` and is in beta. Importing from it raises a `LangChainBetaWarning` once per process. The API may change.

  If you used MCP before v1.4.0, see [Migrate from `langchain-mcp-adapters`](/oss/python/migrate/langchain-mcp-adapters).
</Note>

## Install

Install LangChain with the `mcp` extra, which pulls in FastMCP:

<CodeGroup>
  ```bash pip theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
  pip install "langchain[mcp]"
  ```

  ```bash uv theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
  uv add "langchain[mcp]"
  ```
</CodeGroup>

## Quickstart

Open an [`MCPAdapter`](https://reference.langchain.com/python/langchain/mcp/adapter/MCPAdapter), discover the server's tools with `list_tools()`, and build the agent inside the context. The tools hold the client, so the agent stays usable after the context exits:

```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
from langchain.agents import create_agent
from langchain.mcp import MCPAdapter


async def main():
    async with MCPAdapter("https://example.com/mcp") as adapter:
        tools = await adapter.list_tools()
        agent = create_agent("claude-sonnet-5", tools)
        return await agent.ainvoke({"messages": [{"role": "user", "content": "..."}]})
```

## Transports

[`MCPAdapter`](https://reference.langchain.com/python/langchain/mcp/adapter/MCPAdapter) infers the transport from the target you hand it, so the only thing that changes between an in-process server, a local script over stdio, and a remote URL is the target itself:

```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
from pathlib import Path

from langchain.mcp import MCPAdapter

# An in-process FastMCP server: no subprocess, no socket. Ideal for tests.
in_memory = MCPAdapter(server)  # a FastMCP instance

# A script path is launched over stdio, one subprocess per adapter.
stdio = MCPAdapter(Path("weather_server.py"))

# A string must be an http(s) URL, reached over streamable HTTP.
http = MCPAdapter("https://example.com/mcp")
```

A target can be any of the following:

* **An `http`/`https` URL** (`str`): reached over streamable HTTP.
* **A script path** (`Path`): launched as a subprocess over stdio.
* **A transport object** (`StreamableTransport`): a pre-configured transport object. See [Client Transports](https://gofastmcp.com/clients/transports).
* **An in-process `FastMCP` server**: connected in-memory, with no subprocess or socket.
* **An `MCPConfig` dict** (`{"mcpServers": {...}}`): several servers behind one adapter. See [Connections](/oss/python/langchain/mcp/connections#multiple-servers).
* **A prebuilt `fastmcp.Client`**: for full control over transport, [caching](https://gofastmcp.com/clients/client#response-caching), and [protocol negotiation](https://gofastmcp.com/clients/client#protocol-negotiation).

<Warning>
  A `str` target must be an `http` or `https` URL. FastMCP resolves a string by testing it as a filesystem path before testing it as a URL, so a string naming an existing `.py` or `.js` file would launch that file as a subprocess. Because strings are the form a target most often arrives in from configuration or from a model, [`MCPAdapter`](https://reference.langchain.com/python/langchain/mcp/adapter/MCPAdapter) rejects strings that don't match the shape of a URL.
</Warning>

## Next steps

<CardGroup cols={2}>
  <Card title="Connections" icon="plug" href="/oss/python/langchain/mcp/connections">
    Connection lifecycle, multiple servers, protocol eras, and caching.
  </Card>

  <Card title="Authentication" icon="lock" href="/oss/python/langchain/mcp/auth">
    Bearer tokens, OAuth 2.1, and per-user server auth.
  </Card>

  <Card title="Tools" icon="tool" href="/oss/python/langchain/mcp/tools">
    Load MCP tools into agents, control their execution, and handle their outputs.
  </Card>
</CardGroup>

***

<div className="source-links">
  <Callout icon="terminal-2">
    [Connect these docs](/use-these-docs) to Claude, VSCode, and more via MCP for real-time answers.
  </Callout>

  <Callout icon="edit">
    [Edit this page on GitHub](https://github.com/langchain-ai/docs/edit/main/src/oss/langchain/mcp/index.mdx) or [file an issue](https://github.com/langchain-ai/docs/issues/new/choose).
  </Callout>
</div>