MCP Server

🚀 MCP-Server

This is a server project that serves as the backend for an application. It implements the Model Context Protocol (MCP), enabling AI models like LM Studio to securely interact with local data and custom tools. For example, the server processes data in CSV and Parquet formats and provides tools for data processing and analysis.

🚀 Quick Start

✨ Features

• Implements the Model Context Protocol (MCP).
• Enables AI models to interact with local data and custom tools.
• Processes data in CSV and Parquet formats.
• Provides data processing and analysis tools.

📦 Installation

It is highly recommended to use uv for installing and managing dependencies as it is a fast and modern Python project manager.

1. Install uv: Ensure that uv is installed on your system. If not, you can install it as follows:

   curl -LsSf https://astral.sh/uv/install.sh | sh
   

   Or for Windows (if curl is not available):

   pip install uv
   

   or

   powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
   

   Restart your terminal so that the uv command is available. Check the installation with uv --version.

2. Clone the repository:

   git clone <repository-url>
   cd mcp-server
   

3. Set up the virtual environment and install dependencies:

   uv venv
   .venv\Scripts\activate # On MacOS or Linux: source .venv/bin/activate
   uv pip install -r requirements.txt
   

💻 Usage Examples

Basic Usage

To start the MCP server, navigate to the project root directory and run:

uv run main.py

The server is now waiting for a connection from a client like LM Studio. However, the server does not need to run continuously as LM Studio can start it automatically (in the background) when needed.

Advanced Usage

To connect LM Studio to your MCP server, you need to modify the LM Studio configuration file.

1. Edit mcp.json in LM Studio: Edit the mcp.json file in LM Studio.

2. Add the server to the configuration: Add the following JSON code to the mcp.json file and replace /ABSOLUTE/PATH/TO/mcp-server with the actual absolute path to your project folder. Ensure that uv is in the system path or provide the full path to the uv executable:

   {
     "mcpServers": {
       "mcp-server": {
         "command": "uv",
         "args": [
           "--directory",
           "/ABSOLUTE/PATH/TO/mcp-server",
           "run",
           "main.py"
         ]
       }
     }
   }
   

3. Restart LM Studio: After restarting LM Studio, the tools of your MCP server should be available. Check the LM Studio settings or interface to find the registered tools (summarize_csv_file, summarize_parquet_file).

You can now ask LM Studio to use your tools, for example:

• "Summarize the CSV file named sample.csv."
• "How many rows does sample.parquet have?"

📚 Documentation

Project Structure

• main.py: The main file to start the MCP server.
• server.py: Contains the logic for the MCP server and registers the tools.
• generate_parquet.py: A script to generate Parquet files from other data sources (e.g., sample.csv).
• data/: A directory for sample data (sample.csv, sample.parquet).
• tools/: Contains the MCP tool definitions:
  • csv_tools.py: Functions for processing CSV files.
  • parquet_tools.py: Functions for processing Parquet files.
• utils/: Contains reusable helper functions for reading files:
  • file_reader.py: Helper functions for reading CSV and Parquet files and summarizing their contents.
• pyproject.toml: The project configuration file that manages the project dependencies.
• uv.lock: The dependency lock file generated by uv.

🔧 Technical Details

The project implements the Model Context Protocol (MCP) to enable AI models to interact with local data and custom tools. The server processes data in CSV and Parquet formats and provides data processing and analysis tools. The tools are registered with the MCP server and can be used by AI models like LM Studio.

📄 License

No license information is provided in the original document.

Data

Sample data can be found in the data/ directory.

Tools

The tools in this directory (csv_tools.py, parquet_tools.py) are primarily intended for integration with the MCP server via main.py and server.py. They do not need to be run directly as the MCP server calls them when needed.

Utilities (utils/)

The utils/ directory contains reusable helper functions (e.g., file_reader.py) used by the tools to process and summarize data.

Acknowledgments

This project is inspired by and based on the insights of @llm-guy.

Building a Basic MCP Server with Python