Home / MCP / Kaggle MCP Server

Kaggle MCP Server

Provides access to Kaggle competitions, datasets, kernels, and models via MCP for Claude AI.

other

23stars

View on GitHub

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
    "mcpServers": {
        "kaggle": {
            "command": "kaggle-mcp",
            "args": []
        }
    }
}

The Kaggle MCP server connects Claude AI to the Kaggle API through the Model Context Protocol, enabling you to perform competitions, access datasets, kernels, and pre-trained models directly from the AI interface with secure authentication and streamlined workflows.

How to use

You interact with Kaggle MCP through an MCP client by starting the local Kaggle MCP server and configuring Claude Desktop to connect to it. Once connected, you can authenticate with Kaggle, list active competitions, discover datasets, search kernels, and download data or models for analysis. Use natural language prompts to Claude such as requesting a list of competitions or searching for datasets by topic, and Claude will translate your request into Kaggle API operations via MCP.

How to install

# macOS / Linux
curl -LsSf https://raw.githubusercontent.com/54yyyu/kaggle-mcp/main/install.sh | sh

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

# Manual installation
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

Manual configuration example

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

Configure Claude Desktop to use Kaggle MCP

After you install, set up Claude Desktop to connect to the Kaggle MCP server. The configuration entry places Kaggle MCP as a local stdio server so Claude can invoke the MCP process directly.

Kaggle API credentials

To enable secure access to Kaggle data, complete API credential setup and place the credentials in the expected location. This enables Claude to authenticate with Kaggle through the MCP server.

Available tools

authenticate

Authenticate with Kaggle using your Kaggle API credentials or Claude via the authenticate() workflow to establish a secure session with the MCP server.

list_competitions

Retrieve a list of current Kaggle competitions, including titles, deadlines, and brief descriptions.

download_competition_data

Download data files for a selected competition for offline analysis within Claude.

search_datasets

Search Kaggle datasets by keywords, tags, or topics and download selected items.

download_kernel

Find and retrieve Kaggle kernels/notebooks relevant to your analysis.

list_models

Browse pre-trained models available on Kaggle relevant to your task.