Kaggle-MCP logo

Kaggle-MCP

by 54yyyu

Kaggle-MCP connects Claude AI to the Kaggle API through the Model Context Protocol (MCP), enabling competition, dataset, and kernel operations through the AI interface. It allows users to interact with Kaggle resources directly through Claude AI.

View on GitHub

Last updated: N/A

Kaggle-MCP: Kaggle API Integration for Claude AI

     ██╗  ██╗ █████╗  ██████╗  ██████╗ ██╗     ███████╗       ███╗   ███╗ ██████╗██████╗ 
     ██║ ██╔╝██╔══██╗██╔════╝ ██╔════╝ ██║     ██╔════╝       ████╗ ████║██╔════╝██╔══██╗
     █████╔╝ ███████║██║  ███╗██║  ███╗██║     █████╗         ██╔████╔██║██║     ██████╔╝
     ██╔═██╗ ██╔══██║██║   ██║██║   ██║██║     ██╔══╝  ████─  ██║╚██╔╝██║██║     ██╔═══╝ 
     ██║  ██╗██║  ██║╚██████╔╝╚██████╔╝███████╗███████╗       ██║ ╚═╝ ██║╚██████╗██║     
     ╚═╝  ╚═╝╚═╝  ╚═╝ ╚═════╝  ╚═════╝ ╚══════╝╚══════╝       ╚═╝     ╚═╝ ╚═════╝╚═╝     

Kaggle-MCP connects Claude AI to the Kaggle API through the Model Context Protocol (MCP), enabling competition, dataset, and kernel operations through the AI interface.

Features

  • Authentication: Securely authenticate with your Kaggle credentials
  • Competitions: Browse, search, and download data from Kaggle competitions
  • Datasets: Find, explore, and download datasets from Kaggle
  • Kernels: Search for and analyze Kaggle notebooks/kernels
  • Models: Access pre-trained models available on Kaggle

Quick Installation

The following commands install the base version of Kaggle-MCP.

macOS / Linux

# Install with a single command
curl -LsSf https://raw.githubusercontent.com/54yyyu/kaggle-mcp/main/install.sh | sh

Windows

# Download and run the installer
powershell -c "Invoke-WebRequest -Uri https://raw.githubusercontent.com/54yyyu/kaggle-mcp/main/install.ps1 -OutFile install.ps1; .\install.ps1"

Manual Installation

# Install with pip
pip install git+https://github.com/54yyyu/kaggle-mcp.git

# Or better, install with uv
uv pip install git+https://github.com/54yyyu/kaggle-mcp.git

Configuration

After installation, run the setup utility to configure Claude Desktop:

kaggle-mcp-setup

This will locate and update your Claude Desktop configuration file, which is typically found at:

  • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
  • Windows: %APPDATA%\Claude\claude_desktop_config.json
  • Linux: ~/.config/Claude/claude_desktop_config.json

Manual Configuration

Alternatively, you can manually add the following to your Claude Desktop configuration:

{
  "mcpServers": {
    "kaggle": {
      "command": "kaggle-mcp"
    }
  }
}

Kaggle API Credentials

To use Kaggle-MCP, you need to set up your Kaggle API credentials:

  1. Go to your Kaggle account settings
  2. In the API section, click "Create New API Token"
  3. This will download a kaggle.json file with your credentials
  4. Move this file to ~/.kaggle/kaggle.json (create the directory if needed)
  5. Set the correct permissions: chmod 600 ~/.kaggle/kaggle.json

Alternatively, you can authenticate directly through Claude using the authenticate() tool with your username and API key.

Available Tools

For a comprehensive list of available tools and their detailed usage, please refer to the documentation at stevenyuyy.us/kaggle-mcp.

Examples

Ask Claude:

  • "Authenticate with Kaggle using my username 'username' and key 'apikey'"
  • "List active Kaggle competitions"
  • "Show me the top 10 competitors on the Titanic leaderboard"
  • "Find datasets about climate change"
  • "Download the Boston housing dataset"
  • "Search for kernels about sentiment analysis"

Use Cases

  • Competition Research: Quickly access competition details, data, and leaderboards
  • Dataset Discovery: Find and download datasets for analysis projects
  • Learning Resources: Locate relevant kernels and notebooks for specific topics
  • Model Discovery: Find pre-trained models for various machine learning tasks

Requirements

  • Python 3.8 or newer
  • Claude Desktop or API access
  • Kaggle account with API credentials
  • MCP Python SDK 1.6.0+

License

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