# sunriseapps/imagesorcery-mcp [Health: Active]

**Category:** 🎥 Multimedia Process  
**Repository:** https://github.com/sunriseapps/imagesorcery-mcp  
**GitHub Stars:** 334  
**Views:** 6  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/sunriseapps-imagesorcery-mcp

## Description
ComputerVision-based 🪄 sorcery of image recognition and editing tools for AI assistants.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "imagesorcery-mcp": {
    "command": "uvx",
    "args": ["imagesorcery-mcp"]
  }
}
```

## Documentation

## What sunriseapps/imagesorcery-mcp MCP server does

sunriseapps/imagesorcery-mcp MCP server connects an MCP-compatible AI assistant to local image-processing operations. It can crop, resize, rotate, blur, recolor, overlay, and fill image regions. Drawing tools add text, arrows, circles, lines, and rectangles, which is useful for annotations, redactions, diagrams, and watermarks.

Recognition features include object detection with Ultralytics models, text-based object finding, and OCR through EasyOCR. Detection can produce polygons or segmentation masks as PNG files. The server can also return image metadata and expose the models available in its models directory. A supplied `remove-background` prompt describes a workflow that combines detection and masking to isolate a subject.

All processing is described as local. That makes the server suitable for workflows where source images should remain on the machine running the MCP process.

## How it works

The assistant calls individual MCP tools or combines several tools to complete a larger task. For example, a workflow can find an object in a photo, crop around it, and save the result. Detection accepts a confidence threshold, while OCR accepts language information. Operations use local file paths; the project specifically recommends using full paths in prompts.

The server also provides the `models://list` resource, which reports the models in the local models directory. The `config` tool can show or update ImageSorcery MCP settings, including the default detection confidence mentioned in the examples. The `remove-background` prompt provides structured guidance for a multi-step background-removal task.

## Setup and configuration

sunriseapps/imagesorcery-mcp MCP server requires Python 3.10 or newer and an MCP client such as Claude.app, Cursor, or Cline. The documented installation path uses `pipx`:

```bash
pipx install imagesorcery-mcp
imagesorcery-mcp --post-install
```

The post-install command downloads required models and attempts to install the `clip` Python package from GitHub. OpenCV may also require `ffmpeg`, `libsm6`, `libxext6`, and `libgl1-mesa-glx`; the README provides Ubuntu/Debian and Docker installation commands for those libraries.

For an MCP configuration, the documented stdio entry uses the `imagesorcery-mcp` command and a timeout of 100. Configuration can be inspected or changed through the `config` tool. The README’s Cline instructions also show enabling telemetry through that tool only after requesting user permission.

## Tools and capabilities

The available tools cover:

- Image changes: `crop`, `resize`, `rotate`, `blur`, `change_color`, `fill`, and `overlay`.
- Annotations: `draw_texts`, `draw_arrows`, `draw_circles`, `draw_lines`, and `draw_rectangles`.
- Recognition: `detect`, `find`, and `ocr`.
- Inspection and settings: `get_metainfo` and `config`.

`detect` uses Ultralytics models, while `find` accepts a text description such as an object category. Both can return polygons or segmentation masks. `ocr` uses EasyOCR to extract text from images. The tools generally accept an input image and output path, with operation-specific parameters such as coordinates, colors, opacity, rotation angle, language, or confidence.

## Limitations and notes

Results depend on the selected recognition model and confidence setting. Required system libraries can be missing in minimal environments such as Docker containers. Detailed parameter references are maintained in the repository’s tool, resource, and prompt README files.

_Full upstream README: https://allmcps.com/mcp/sunriseapps-imagesorcery-mcp/readme_