1sarthakbhardwaj/labellerr-mcp-server

Labellerr MCP Server

A Model Context Protocol (MCP) server that provides a comprehensive interface to the Labellerr SDK for managing annotation projects, datasets, and monitoring operations through AI assistants like Claude Desktop and Cursor.

Features

• 🚀 Project Management - Create, list, update, and track annotation projects
• 📊 Dataset Management - Create datasets, upload files/folders, and query information
• 🏷️ Annotation Tools - Upload pre-annotations, export data, and download results
• 📈 Monitoring & Insights - Real-time progress tracking and system health monitoring
• 🔍 Query Capabilities - Search projects, get statistics, and analyze operations

22 specialized tools available across 5 categories to streamline your annotation workflow.

Installation

Prerequisites

• Node.js 16 or higher
• npm or yarn
• Labellerr API credentials (API Key, API Secret, Client ID)

Setup

1. Clone the repository:

git clone https://github.com/1sarthakbhardwaj/labellerr-mcp-server.git
cd labellerr-mcp-server

2. Install dependencies:

npm install

3. Configure environment variables:

cp .env.example .env

Edit .env and add your Labellerr credentials:

LABELLERR_API_KEY=your_api_key_here
LABELLERR_API_SECRET=your_api_secret_here
LABELLERR_CLIENT_ID=your_client_id_here

Getting Credentials: Contact Labellerr support or email support@labellerr.com to obtain your API credentials.

Configuration

Option 1: Using with Claude Desktop

Add to your Claude Desktop configuration file:

Location: ~/Library/Application Support/Claude/claude_desktop_config.json (macOS)

{
  "mcpServers": {
    "labellerr": {
      "command": "node",
      "args": ["/absolute/path/to/labellerr-mcp-server/src/index.js"],
      "env": {
        "LABELLERR_API_KEY": "your_api_key",
        "LABELLERR_API_SECRET": "your_api_secret",
        "LABELLERR_CLIENT_ID": "your_client_id"
      }
    }
  }
}

Important: Replace /absolute/path/to/ with the full path to your installation directory.

After configuration:

1. Restart Claude Desktop completely
2. The Labellerr tools will be available in your conversations
3. Ask Claude to list your projects or check system health

Option 2: Using with Cursor

Add to your Cursor MCP configuration file:

Location: ~/.cursor/mcp.json (macOS/Linux) or %APPDATA%\Cursor\mcp.json (Windows)

{
  "mcpServers": {
    "labellerr": {
      "command": "node",
      "args": ["/absolute/path/to/labellerr-mcp-server/src/index.js"],
      "env": {
        "LABELLERR_API_KEY": "your_api_key",
        "LABELLERR_API_SECRET": "your_api_secret",
        "LABELLERR_CLIENT_ID": "your_client_id"
      }
    }
  }
}

Important: Replace /absolute/path/to/ with the full path to your installation directory.

After configuration:

1. Restart Cursor completely (Quit and reopen)
2. The Labellerr tools will be available in the AI assistant
3. Try asking: "List all my Labellerr projects"

Verifying Installation

Test the server is working:

# Start the server
npm start

# In another terminal, test the protocol
echo '{"jsonrpc":"2.0","method":"tools/list","id":1}' | node src/index.js

You should see a JSON response listing all 22 available tools.

Usage

Starting the Server Standalone

# Production mode
npm start

# Development mode (with auto-reload)
npm run dev

Using with AI Assistants

Once configured with Claude Desktop or Cursor, you can interact naturally:

Project Management:

• "List all my Labellerr projects"
• "Create a new image classification project for product categorization"
• "What's the progress of project XYZ?"

Dataset Operations:

• "Upload images from /path/to/folder"
• "List all my datasets"
• "Create a new dataset for video annotation"

Monitoring:

• "Show me system health"
• "Check the progress of my active projects"
• "What operations have been performed?"

Exports:

• "Export annotations in COCO format"
• "Check status of export ABC123"
• "Download completed export"

Current Status

✅ Fully Working (21 tools)

• Project Management: List, get details, update rotation
• Dataset Management: Create, upload, list, query
• Annotation Operations: Upload pre-annotations, export, download
• Monitoring: Job status, progress, system health
• Query & Search: Statistics, history, search

⚠️ In Progress (1 tool)

• Project Creation - Implementation complete but encountering API 400 error during dataset creation
  • File upload to GCS: ✅ Implemented
  • Dataset creation: ⚠️ Getting 400 error
  • Template creation: ✅ Implemented
  • Project finalization: ✅ Implemented
  • See Issue #1 for details

Available Tools

The server provides 22 specialized tools:

📋 Project Management (4 tools)

• project_create - Create projects with annotation guidelines
• project_list - List all projects
• project_get - Get detailed project information
• project_update_rotation - Update rotation configuration

📊 Dataset Management (5 tools)

• dataset_create - Create new datasets
• dataset_upload_files - Upload individual files
• dataset_upload_folder - Upload entire folders
• dataset_list - List all datasets
• dataset_get - Get dataset information

🏷️ Annotation Operations (5 tools)

• annotation_upload_preannotations - Upload pre-annotations (sync)
• annotation_upload_preannotations_async - Upload pre-annotations (async)
• annotation_export - Create annotation export
• annotation_check_export_status - Check export status
• annotation_download_export - Get export download URL

📈 Monitoring & Analytics (4 tools)

• monitor_job_status - Monitor background job status
• monitor_project_progress - Track project progress
• monitor_active_operations - List active operations
• monitor_system_health - Check system health

🔍 Query & Search (4 tools)

• query_project_statistics - Get detailed project stats
• query_dataset_info - Get dataset information
• query_operation_history - View operation history
• query_search_projects - Search projects by name/type

For detailed parameters and examples, see the Full Tool Documentation below.

Supported Data Types

• image - JPEG, PNG, TIFF
• video - MP4
• audio - MP3, WAV
• document - PDF
• text - TXT

Annotation Types

• BoundingBox - Rectangle annotations for object detection
• polygon - Polygon shapes for segmentation
• dot - Point annotations
• radio - Single choice selection
• dropdown - Dropdown selection
• boolean - Yes/No selection
• input - Text input field
• select - Multiple choice selection

Export Formats

• json - Standard JSON format
• coco_json - COCO dataset format
• csv - Comma-separated values
• png - Image masks

Limits

• Maximum 2,500 files per folder upload
• Maximum 2.5 GB total folder size
• Batch processing: 15 MB per batch, 900 files max

Example Workflows

1. Create an Object Detection Project

{
  "project_name": "Vehicle Detection",
  "dataset_name": "Traffic Dataset",
  "data_type": "image",
  "created_by": "user@example.com",
  "annotation_guide": [
    {
      "question": "Detect Vehicles",
      "option_type": "BoundingBox",
      "required": true,
      "options": [{"option_name": "#ff0000"}]
    }
  ],
  "folder_to_upload": "/path/to/images"
}

2. Monitor Project Progress

Ask your AI assistant: "Show me the progress of my annotation projects"

The server will return:

• Total files
• Annotated count
• Reviewed count
• Completion percentage

3. Export Annotations

{
  "project_id": "proj_abc123",
  "export_name": "Training Export",
  "export_format": "coco_json",
  "statuses": ["accepted", "reviewed"]
}

4. Search Projects

Ask: "Find all projects related to 'vehicle' or 'traffic'"

The server will search project names and return matching results.

Detailed Tool Reference

Troubleshooting

Server won't start

• Verify Node.js version (requires 16+)
• Check environment variables are set correctly
• Ensure port is not in use

Tools return errors

• Verify Labellerr API credentials are correct
• Check network connectivity
• Review operation history for error details

AI assistant can't find tools

• Verify configuration file path is correct
• Use absolute paths, not relative paths
• Restart the AI assistant completely after configuration
• Check that credentials are set in the config file

Debug Mode

Set LOG_LEVEL=debug in your .env file for detailed logging.

Development

Project Structure

labellerr-mcp-server/
├── src/
│   ├── index.js              # Main server entry point
│   ├── labellerr-client.js   # Labellerr API client
│   └── tools/
│       └── index.js          # Tool definitions
├── package.json              # Dependencies and scripts
├── .env.example              # Environment template
├── claude_desktop_config.json # Claude configuration example
├── LICENSE                   # MIT License
└── README.md                 # This file

Adding New Tools

1. Define the tool schema in src/tools/index.js
2. Implement the handler in src/index.js (handleCallTool method)
3. Add the client method in src/labellerr-client.js if needed
4. Update documentation

Resources

• Labellerr Documentation: docs.labellerr.com
• MCP Protocol: modelcontextprotocol.io
• Support Email: support@labellerr.com
• GitHub Issues: github.com/1sarthakbhardwaj/labellerr-mcp-server/issues

License

MIT License - see file for details.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Made with ❤️ for the Labellerr community