Happy slice of mcp pizzaMCP.pizza

opencv-mcp-server

MCP.Pizza Chef: GongRzhe

Twenty-two commands cover the everyday work — resize, crop, change color mode, blur, outline edges and shapes, report image statistics — plus finding faces, matching features between pictures, pulling frames out of a video, spotting motion and tracking a moving object. Files stay on your computer and no account or credential is involved. Two things to know before you commit: the author archived the repository in September 2025, so it is frozen and read-only, and every command that produces a picture also opens it in your desktop viewer.

Archived · The author made this repository read-only on GitHub.
Files/PDF
Images/Design

Use This MCP server To

Resize or crop a batch of photos by asking Find the faces in a photo and box them Pull still frames out of a video file Spot where movement happens in a security clip Check the brightness and color spread of an image Follow one object across a video clip

README

logo

🚀 OpenCV MCP Server

MCP server providing powerful OpenCV computer vision capabilities for AI assistants.

🔍 Introduction

OpenCV MCP Server is a Python package that provides OpenCV's image and video processing capabilities through the Model Context Protocol (MCP). This allows AI assistants and language models to access powerful computer vision tools for tasks ranging from basic image manipulation to advanced object detection and tracking.

With OpenCV MCP Server, AI systems can:

  • Process and analyze images in various formats
  • Perform real-time object detection and tracking
  • Extract meaningful information from visual data
  • Enhance and transform images using advanced algorithms
  • Process video content with frame-by-frame analysis

✨ Features

  • 📸 Basic image handling and manipulation (read, save, convert)
  • 🖼️ Image processing and enhancement (resize, crop, filter application)
  • 📊 Edge detection and contour analysis
  • 🧠 Advanced computer vision capabilities (feature detection, object detection)
  • 😃 Face detection and recognition
  • 🎬 Video processing and analysis (frame extraction, motion detection)
  • 🔍 Object tracking in videos
  • 📹 Camera integration for real-time object detection

📊 Demo Examples

opencv-mcp-server FAQ

Is this still maintained?
No. The author archived the repository in September 2025, so it is read-only — no fixes, no new features, and open issues stay open. It still runs, but nobody is looking after it.
Do I need an account or key?
No. Everything runs on your own machine against files you already have, and no credential is read anywhere in the code.
Can I use this to spot objects in a photo, like cars or dogs?
Only after a manual download. The instructions claim the detection model is already included, but the 237 MB model file is not in the project — until you fetch it yourself, that command replies with download steps instead of results.
Does face detection work straight away?
Yes, in its default mode, because it borrows the face pattern file that comes with OpenCV. It finds where faces are; it does not tell you who they are, despite the word 'recognition' on the project page.
What happens to the pictures it makes?
Each one is saved next to your original with a timestamped name and then opened in your computer's default viewer, so expect new windows and extra files to pile up as you work.
Can I choose where a file gets saved?
Not really. The save command accepts an output path but then writes to its own timestamped name in that folder instead, so the exact filename you ask for is ignored.
Which apps does it work with?
Claude Desktop is the documented one, using a short entry in its settings file. Anything else that accepts a standard MCP entry should work as well.
Do all the listed settings work?
No. The folder setting for face patterns, CV_HAAR_CASCADE_DIR, is never actually read — it only appears inside an error message. The folder setting for detection models is read and does work.
Can it use my webcam?
Yes, one command switches on the camera to look for objects live. It needs the same missing model file first, and it turns the camera on as soon as it is asked.
Will it overwrite my original photos?
No. It always writes a new file rather than replacing the original, though there is no preview or undo, so work on copies of anything irreplaceable.