home / mcp / kaggle mcp server

Kaggle MCP Server

Provides an MCP server to interact with Kaggle from Claude Desktop using MCP clients.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "krishnapramodparupudi-kaggle-mcp-server": {
      "command": "python",
      "args": [
        "kaggle-server.py"
      ]
    }
  }
}

You can run a dedicated MCP server that lets Claude Desktop or any MCP-compatible client interact with Kaggle. This server exposes Kaggle API methods through MCP tools, enabling you to fetch competitions and other Kaggle data seamlessly from your MCP client.

How to use

To use the Kaggle MCP server, first ensure you have the server running locally. Then configure your MCP client (Claude Desktop or equivalent) to connect to this MCP server and access the available tools.

Start the Kaggle MCP server by running the Python script responsible for the MCP interface. You will connect via MCP tooling to call functions such as retrieving the list of Kaggle competitions. For example, you can invoke the available tool to fetch competitions and receive a JSON-friendly response you can use in your workflows.

Once connected, you can use the server’s tools to perform actions like listing competitions. The tool is designed to return structured results that you can easily integrate into your MCP-driven automation or chat-based workflows.

How to install

# Prerequisites
python3 --version
python3 -m pip --version

# 1) Install dependencies
pip install -r requirements.txt

# 2) Prepare Kaggle credentials
# Place kaggle.json in the appropriate location as described below
# Linux/Mac: ~/.kaggle/kaggle.json
# Windows: C:\Users\<your-username>\.kaggle\kaggle.json

# 3) Configure Claude Desktop (MCP client) to connect to this server
# Ensure Claude/Desktop config includes the Kaggle MCP server entry
# Example content for Claude/claude_desktop_config.json is shown in the guide

# 4) Run the MCP server
# Start Claude Desktop, then start the Kaggle MCP server
# Run the server script as a Python file to initialize the MCP interface
# Command can be executed from the repository path where kaggle-server.py resides
python kaggle-server.py

Additional content

Configuration notes ensure you do not publicly expose your Kaggle credentials. Never commit your kaggle.json file to version control, and restrict access to the .kaggle folder. The credentials enable the server to authenticate with Kaggle on your behalf.

Security considerations include keeping your Kaggle credentials secure and ensuring the MCP server is only accessible from trusted clients. If you work in shared environments, use proper file permissions and network controls to limit exposure.

Tools exposed by the server include get_competitions_list, which fetches available Kaggle competitions and returns them in a JSON-friendly format for easy consumption in MCP workflows.

Available tools

get_competitions_list

Fetches the list of available Kaggle competitions and returns them in a JSON-friendly format.