langchain-box-mcp-adapter

What is langchain-box-mcp-adapter?

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

How to use langchain-box-mcp-adapter?

Clone the repository, install dependencies with uv sync, create a .env file with the required credentials (LangSmith, OpenAI, Box), and ensure the Box MCP server is accessible. Update the StdioServerParameters path in src/simple_client.py or src/graph.py with the correct path to your MCP server. Run the simple client with uv run src/simple_client.py or the graph-based agent via langgraph dev.

Key features of langchain-box-mcp-adapter

• Langchain integration using ChatOpenAI model.
• Connects to Box MCP server via stdio transport.
• Dynamically loads tools from the MCP server.
• Creates a React-style agent for user prompts.
• Provides rich console output with markdown and typewriter effects.

Use cases of langchain-box-mcp-adapter

• Interact with Box data through an AI agent in a console application.
• Build complex workflows using the LangGraph-based agent.
• Demonstrate the Langchain-MCP adapter pattern for Box integration.

FAQ from langchain-box-mcp-adapter

What dependencies are required?

Python 3.13 or higher, and packages: langchain-mcp-adapters>=0.0.8, langchain-openai>=0.3.12, langgraph>=0.3.29, and rich>=14.0.0.

How is the connection to Box MCP set up?

The adapter uses stdio transport with StdioServerParameters pointing to the Box MCP server script, configured via command uv.

What credentials are needed?

API keys for LangSmith and OpenAI, plus Box client ID, client secret, subject type, and subject ID, all placed in a .env file.

Can I run a graph-based agent?

Yes, by invoking the make_graph function in src/graph.py and running langgraph dev with the provided configuration.

Is there a license for this project?

Yes, the project is licensed under the MIT License.