Kaggle MCP Server

What is Kaggle MCP Server?

A Model Context Protocol (MCP) server that enables AI assistants like Claude to interact with Kaggle competitions, including listing competitions, downloading files, submitting predictions, and viewing submissions. Designed for data scientists and AI users who want to manage Kaggle tasks through a conversational interface.

How to use Kaggle MCP Server?

Install Node.js 16+, TypeScript, and the Kaggle CLI (pip install kaggle). Clone the repo, install dependencies (npm install), build (npm run build), and start (npm start). Configure Kaggle API credentials either via a kaggle.json file or environment variables in your claude_desktop_config.json. Then connect with Claude Desktop by adding the server configuration to the same config file.

Key features of Kaggle MCP Server

• List active Kaggle competitions with pagination and search
• Download competition files to a custom path
• Submit prediction files with custom messages
• View submission history, status, and scores
• Supports dataset browsing, notebook integration, and user management (feature-complete branch)

Use cases of Kaggle MCP Server

• Browse ongoing Kaggle competitions and filter by keyword
• Download competition datasets for offline analysis
• Submit model predictions to a competition from the command line
• Track submission history and performance over time

FAQ from Kaggle MCP Server

How do I authenticate with the Kaggle API?

Use either a kaggle.json file placed in ~/.kaggle/ (Linux/Mac) or C:\Users\<USERNAME>\.kaggle\ (Windows), or set the KAGGLE_USERNAME and KAGGLE_KEY environment variables in your claude_desktop_config.json.

What are the prerequisites for using the server?

Node.js 16 or higher, TypeScript, the Kaggle CLI (pip install kaggle), and valid Kaggle API credentials.

Where should the kaggle.json file be placed?

• Linux/Mac: ~/.kaggle/kaggle.json
• Windows: C:\Users\<USERNAME>\.kaggle\kaggle.json
  The file must have permissions 600 on Linux/Mac.

How do I verify the Kaggle CLI is installed?

Run kaggle --version in your terminal. If it is not found, ensure pip install kaggle completed successfully and the command is in your PATH.

What should I do if I receive authentication errors?

Verify credentials are correctly configured, check the kaggle.json location and permissions, and generate a new API token from your Kaggle account if needed.