langchain-box-mcp-adapter

This sample project implements the Langchain MCP adapter to the Box MCP server. It demonstrates how to integrate Langchain with a Box MCP server using tools and agents.

Features

• Langchain Integration: Utilizes Langchain's ChatOpenAI model for AI interactions.
• MCP Server Communication: Connects to the Box MCP server using stdio transport.
• Tool Loading: Dynamically loads tools from the MCP server.
• Agent Creation: Creates a React-style agent for handling user prompts and tool interactions.
• Rich Console Output: Provides a user-friendly console interface with markdown rendering and typewriter effects.

Requirements

• Python 3.13 or higher
• Dependencies listed in pyproject.toml:
  • langchain-mcp-adapters>=0.0.8
  • langchain-openai>=0.3.12
  • langgraph>=0.3.29
  • rich>=14.0.0

Setup

1. Clone the repository:

   git clone <repository-url>
   cd langchain-box-mcp-adapter
   

2. Install dependencies:

   uv sync
   

3. Create a .env file in the root of the project and fill in the information.

LANGSMITH_TRACING = "true"
LANGSMITH_API_KEY =
OPENAI_API_KEY =

BOX_CLIENT_ID = ""
BOX_CLIENT_SECRET = ""
BOX_SUBJECT_TYPE = "user"
BOX_SUBJECT_ID = ""

3. Ensure the MCP server is set up and accessible at the specified path in the project.

4. Update the StdioServerParameters in src/simple_client.py or src/graph.py with the correct path to your MCP server script.

    server_params = StdioServerParameters(
        command="uv",
        args=[
            "--directory",
            "/your/absolute/path/to/the/mcp/server/mcp-server-box",
            "run",
            "src/mcp_server_box.py",
        ],
    )

Usage

Running the Simple Client

To run the simple client:

uv run src/simple_client.py

This will start a console-based application where you can interact with the AI agent. Enter prompts, and the agent will respond using tools and AI capabilities.

Running the Graph-Based Agent (LangGraph)

The graph-based agent can be used by invoking the make_graph function in src/graph.py. This is useful for more complex workflows.

uv run langgraph dev --config src/langgraph.json

You should see something like:

Project Structure

• src/simple_client.py: Main entry point for the simple client.
• src/graph.py: Contains the graph-based agent setup.
• src/console_utils/console_app.py: Utility functions for console interactions.
• src/langgraph.json: Configuration for the LangGraph integration.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Contributing

Contributions are welcome! Please open an issue or submit a pull request for any improvements or bug fixes.