# mcp-kaggle-tool

MCP server for Kaggle API integration - create, run, and manage Kaggle notebooks programmatically.

## 🚀 Features

- ✅ Authentication check for Kaggle API
- 📝 Create and manage Kaggle notebooks
- 🏃 Run notebooks with GPU support
- 📊 Search datasets and competitions
- 💾 Download notebook outputs
- 🔍 Monitor execution status

## 📋 Prerequisites

1. **Kaggle Account**: You need a Kaggle account
2. **Kaggle API Token**: 
   - Go to https://www.kaggle.com/account
   - Click "Create New API Token"
   - Save the downloaded `kaggle.json` to `~/.kaggle/`
3. **Kaggle CLI**: Install the Kaggle CLI:
   ```bash
   pip install kaggle
   ```

## 🛠️ Installation

### From npm (when published)
```bash
npm install -g mcp-kaggle-tool
```

### From source
```bash
git clone https://github.com/yourusername/mcp-kaggle-tool.git
cd mcp-kaggle-tool
npm install
npm run build
```

## 🔧 Configuration

### For Claude Desktop

Add to your Claude Desktop configuration (`~/Library/Application Support/Claude/claude_desktop_config.json` on macOS):

```json
{
  "mcpServers": {
    "kaggle": {
      "command": "npx",
      "args": ["mcp-kaggle-tool"]
    }
  }
}
```

Or if running from source:

```json
{
  "mcpServers": {
    "kaggle": {
      "command": "node",
      "args": ["/path/to/mcp-kaggle-tool/dist/index.js"]
    }
  }
}
```

## 📚 Available Tools

### Authentication
- `kaggle_auth_check` - Verify Kaggle API credentials are configured

### Notebooks
- `kaggle_list_notebooks` - List your Kaggle notebooks
- `kaggle_create_notebook` - Create a new notebook with code
- `kaggle_run_notebook` - Execute a notebook
- `kaggle_get_notebook_status` - Check execution status
- `kaggle_download_notebook_output` - Download notebook outputs

### Data & Competitions
- `kaggle_search_datasets` - Search for datasets
- `kaggle_list_competitions` - List active competitions

## 💡 Usage Examples

### Check Authentication
```
Use kaggle_auth_check to verify your credentials are set up
```

### Create and Run a Notebook
```
1. Create a notebook with kaggle_create_notebook:
   - title: "My ARC Experiment"
   - code: "print('Hello from Kaggle!')"
   - enableGpu: true

2. Monitor with kaggle_get_notebook_status
3. Download results with kaggle_download_notebook_output
```

### Search ARC Dataset
```
Use kaggle_search_datasets with search: "abstraction reasoning corpus"
```

## 🚧 Development

```bash
# Install dependencies
npm install

# Build TypeScript
npm run build

# Run in development
npm run dev

# Run tests
npm test

# Lint code
npm run lint
```

## 📝 License

MIT License - see LICENSE file for details.

## 🤝 Contributing

Contributions welcome! Please open an issue or submit a PR.

## 🐛 Known Issues

- Kaggle API sometimes returns HTML instead of JSON for certain commands
- Notebook execution status may take time to update
- GPU availability depends on Kaggle quota

## 🔗 Resources

- [Kaggle API Documentation](https://github.com/Kaggle/kaggle-api)
- [MCP Documentation](https://modelcontextprotocol.io)
- [Kaggle Notebooks Guide](https://www.kaggle.com/docs/notebooks)