Pandas MCP Server

This repository contains a server implementation using the Model Context Protocol (MCP) with functionalities to handle CSV files and execute Pandas code.

Requirements

• Python 3.11 or higher
• Install required packages: pip install -r requirements.txt

Functions

load_csv_tool

• Description: Loads a CSV file and returns its column structure and sample data.
• Parameters:
  • file_path: Path to the CSV file.
• Returns:
  • A dictionary containing the columns and sample data from the CSV file.
• Notes:
  • Detects file encoding and delimiter automatically.
  • Limits file size to 100MB to prevent excessive memory usage.

run_pandas_code

• Description: Executes Pandas code provided as a string.
• Parameters:
  • file_path: Path to the CSV file to be loaded into a DataFrame.
  • code: String containing the Pandas code to execute.
• Returns:
  • A dictionary containing the result of the executed code and any variables created during execution.
• Security Notes:
  • Prevents execution of blacklisted operations such as os., sys., subprocess., open(, exec(, eval(, import os, import sys.
  • Provides detailed error messages and suggestions to help users resolve issues.

Usage

1. Configure your MCP client with the following settings:

{
  "mcpServers": {
    "pandas": {
      "name": "pandas",
      "type": "stdio",
      "description": "run pandas code",
      "isActive": true,
      "command": "python",
      "args": [
        "${workspaceFolder}/server.py"
      ]
    }
  }
}

2. Use the configured MCP client to interact with the server and utilize the provided tools.

Workflow

1. Load and inspect your CSV file:

   • User prompt: "Load the CSV file at data/sample.csv and show me the column structure"
   • This will call load_csv_tool with the file path

2. Execute Pandas operations on the loaded data:

   • User prompt: "Group the data by category and calculate the sum for each group"
   • This will call run_pandas_code with the appropriate Pandas operation