Groundlight MCP Server

🚀 Groundlight MCP Server

A Model Context Protocol (MCP) server that interacts with Groundlight. This server provides tools for creating and listing Detectors and ImageQueries.

This MCP server is still in the early stages of development. Functionality and services may change and expand as we continue to develop and improve the server.

🚀 Quick Start

Tools

The following tools are available in the Groundlight MCP server:

1. create_detector

   • Description: Create a detector based on the specified configuration. Three modes are supported:

     1. Binary: Answer "yes" or "no" to a natural language query about an image.
     2. Multi-class: Classify an image into predefined categories based on a natural language query.
     3. Count: Use natural language to describe and count the number of specified objects in an image.

     All detectors analyze images to answer natural language queries and return a confidence score indicating the reliability of the result. If the confidence is below the specified threshold, the query will be escalated to human review. Detectors continuously improve by learning from feedback and additional examples.

   • Input: config (a DetectorConfig object containing name, query, confidence_threshold, mode, and mode-specific configuration)

   • Return: A Detector object

2. get_detector

   • Description: Get a detector by its ID.
   • Input: detector_id (string)
   • Return: A Detector object

3. list_detectors

   • Description: List all detectors associated with the current user.
   • Input: None
   • Return: A list of Detector objects

4. submit_image_query

   • Description: Submit an image to be answered using a specified detector. The image can be provided as a file path, URL, or raw bytes. The detector will return a response with a label and a confidence score.
   • Input: detector_id (string), image (string or bytes)
   • Return: An ImageQuery object

5. get_image_query

   • Description: Get an existing image query by its ID.
   • Input: image_query_id (string)
   • Return: An ImageQuery object

6. list_image_queries

   • Description: List all image queries associated with a specified detector. Note that this may return a large number of results.
   • Input: detector_id (string)
   • Return: A list of ImageQuery objects

7. get_image

   • Description: Get an image by its image query ID. Optionally, annotate the image with bounding boxes.
   • Input: image_query_id (string), annotate (boolean, default is false)
   • Return: An Image object

📚 Documentation

Configuration

Using with Claude Desktop

Add the following to claude_desktop_config.json:

Docker

"mcpServers": {
  "groundlight": {
    "command": "docker",
    "args": ["run", "--rm", "-i", "-e", "GROUNDLIGHT_API_TOKEN", "groundlight/groundlight-mcp-server"],
    "env": {
        "GROUNDLIGHT_API_TOKEN": "YOUR_API_TOKEN_HERE"
    }
  }
}

Development

Build the Docker image locally:

make build-docker

Run the Docker image locally:

make run-docker

[Groundlight Internal] Push the Docker image to Docker Hub (requires DockerHub credentials):

make push-docker