Groundlight MCP Server

Example Use Cases

• 1

  Create, configure, and manage visual detectors (binary, multiclass, counting) for automated image understanding and decisioning.

• 2

  Submit images to detectors and retrieve labeled responses, confidence scores, annotated images, and query histories for auditing or downstream processing.

• 3

  Set up alerts (webhook, email, SMS) based on detector conditions and manage human-in-the-loop labeling for model improvement.

Description

This Model Context Protocol (MCP) server provides tools to create binary, multiclass, and counting detectors that answer natural-language image queries, submit and list image queries, retrieve annotated images and evaluation metrics, and create alerting rules. It supports confidence-based escalation to human review and continuous learning via user labels. The server is distributed as a Docker image and integrates with tools like Claude Desktop and Zed via a provided connector configuration. It is intended to streamline building and operating computer-vision workflows on the Groundlight platform.