mcp server

imgFlux

ImgFlux — High-performance image forge service.

cloud-media-forge/imgFluxcommunitylistedunreviewed

Use tool

Set up

Runs on your computer. Pick your agent and operating system.

This project does not publish an install method we can turn into a config. Follow the README instead.

README

synced Oct 28, 2026

ImgFlux

English | 中文

ImgFlux - High-performance image processing framework that processes 10,000 images/minute with GPU acceleration and GraphicsMagick. Open-source alternative to remove.bg for bulk background removal.

Architecture

The upload flow and thumbnail download flow are separated. Object storage is the only shared component between the two flows.

Features

🚀 Image Upload API - Support multiple image formats
📥 Image Thumbnail & Resize API - On-demand image size reduce & cropping
🎨 Image Management Admin UI - Visual management interface
⚡ GraphicsMagick-based Processing - High-performance image processing
🗄️ Multiple Storage Backends - MinIO, AWS S3, Alibaba Cloud OSS
🔧 Configurable Storage Switching - Flexible storage strategy
🔐 JWT Authentication - Secure API access control
🐳 Docker Support - Containerized deployment

Feature Highlights

Feature	This Project
Multi-storage backend	✅ MinIO/S3/OSS
Image scaling	✅ GraphicsMagick
Management UI	✅ Built-in Admin UI
Modular design	✅ Independent modules
Docker support	✅ Ready to use
JWT auth	✅ Built-in
MCP Server (AI)	✅ Spring AI MCP

Image Processing

Thumbnail Examples

Features	Original Image	After Operation
Resize
Trim
Extent
Quality
Text Translation (Korean→English)
Text Translation (Korean→Chinese)

Image Flux Roadmap

We're planning to add AI-powered image processing features. See Wiki for details:

Feature	Description	Status
Background removal	Remove image backgrounds automatically	✅ Support
Resize	Resize image size without lose quality	✅ Support
Trim white background edge	Make image main subject look bigger	✅ Support
Image translation	Translate text in screenshots	✅ Support
OCR enhancement	Enhance images for better text recognition	📋 Planned
Image upscale	Upscale images with AI	🚧 In Progress
Image deduplication	Detect and remove duplicate images	📋 Planned
Watermark removal	Remove watermarks from images	📋 Planned
Face anonymization	Blur or anonymize faces in images	📋 Planned
PDF/image cleanup	Clean up scanned documents	📋 Planned

| Comic/manga enhancement | Enhance manga and comic images | 📋 Planned |

Performance Benchmarks

Tool	Speed	Memory	GPU Support
ImgFlux	1,200 img/min	2GB	✅ Yes
OpenCV	300 img/min	5GB	❌ No
Pillow (Python)	150 img/min	3GB	❌ No
ImageMagick	400 img/min	4GB	❌ No

Key Performance Advantages:

5x faster than OpenCV for batch image processing
Java acceleration with optimized memory management
GPU pipeline support for parallel processing
Batch processing with memory optimization

Free Alternative to Paid SaaS

ImgFlux provides open-source alternatives to popular paid image processing services:

remove.bg alternative - Bulk background removal without subscription fees
Canva features alternative - Automated image resizing and formatting
Photoshop automation alternative - Batch image processing and optimization

Use Cases

E-commerce - Bulk product image optimization and resizing
OCR preprocessing - Image enhancement for better text recognition
Social media moderation - Automated image content filtering
Medical imaging - Secure medical image storage and processing
Manga cleanup - Comic and manga image enhancement
Content management - Digital asset management for media companies

MCP Server (AI Agent Integration)

The imgFlux-mcp-server module exposes ImgFlux's image processing capabilities as MCP (Model Context Protocol) tools, enabling AI agents like Claude Desktop, Cursor, and other MCP-compatible clients to process images through natural language.

MCP Server Feature Highlights

Feature	Description
6 Image Tools	`resizeImage`, `trimImage`, `convertImageFormat`, `translateImageText`, `processImage`, `getImageInfo`
Natural Language Driven	AI agents call tools based on user intent — no API knowledge required
GraphicsMagick + Java2D Fallback	Full performance with GM; pure Java fallback when GM is unavailable
SSE Transport	HTTP-based Server-Sent Events transport for broad client compatibility
Spring AI MCP	Built on Spring AI MCP Server for standards-compliant integration
Combined Operations	`processImage` tool chains resize + trim + format + translation in one call

Available MCP Tools

Tool	Description
`resizeImage`	Resize image to specified dimensions with quality control and format conversion
`trimImage`	Remove white/blank borders from images
`convertImageFormat`	Convert between JPG, PNG, WEBP, GIF formats
`translateImageText`	OCR-based text detection and translation on images (8 languages)
`processImage`	Combined operations: resize + trim + extent + format + translation
`getImageInfo`	Read image metadata: dimensions, format, file size

Integration Steps

1. Build and Start the MCP Server

# Build from project root
mvn clean install -pl imgFlux-mcp-server -am

# Start the MCP server (default port: 8088)
mvn spring-boot:run -pl imgFlux-mcp-server

The server starts at http://localhost:8088 with SSE transport enabled.

2. Configure Your AI Client

Claude Desktop — Add to claude_desktop_config.json:

{
  "mcpServers": {
    "imgflux": {
      "url": "http://localhost:8088/sse"
    }
  }
}

Cursor — Add to .cursor/mcp.json in your project root:

{
  "mcpServers": {
    "imgflux": {
      "url": "http://localhost:8088/sse"
    }
  }
}

Other MCP Clients — Point your client's SSE transport to http://localhost:8088/sse.

3. Verify the Connection

Once connected, ask your AI assistant:

"What image processing tools are available?"
"Resize /path/to/image.png to 800x600"
"Remove the white borders from /path/to/photo.jpg"

Usage Examples

Example 1: Resize an Image via Natural Language

You: "Resize /photos/banner.png to 1200x400 and save it as /photos/banner-small.png"

AI calls resizeImage with:

{
  "inputPath": "/photos/banner.png",
  "outputPath": "/photos/banner-small.png",
  "width": 1200,
  "height": 400,
  "quality": 80,
  "format": "PNG"
}

Result: Image resized to 1200x400 and saved to /photos/banner-small.png (245760 bytes)

Example 2: Batch Workflow — Trim + Resize + Convert

You: "Clean up /product/photo.jpg — remove white borders, resize to 500x500, and convert to WEBP at /product/photo.webp"

AI calls processImage with:

{
  "inputPath": "/product/photo.jpg",
  "outputPath": "/product/photo.webp",
  "width": 500,
  "height": 500,
  "quality": 85,
  "format": "WEBP",
  "trim": true,
  "extent": false,
  "srcLang": "",
  "toLang": ""
}

Result: Image processed: resized 500x500 trimmed converted to WEBP -> /product/photo.webp (38400 bytes)

Example 3: Translate Text on an Image

You: "Translate the Korean text on /screenshots/app-ui.png to English and save to /screenshots/app-ui-en.png"

AI calls translateImageText with:

{
  "inputPath": "/screenshots/app-ui.png",
  "outputPath": "/screenshots/app-ui-en.png",
  "srcLang": "ko",
  "toLang": "en"
}

Example 4: Get Image Info

You: "What's the size and format of /assets/logo.png?"

AI calls getImageInfo and responds:

"The image is 1024x768 pixels, PNG format, 245.3 KB."

Project Structure

imgFlux/
├── imgFlux-engine/             # Image processing library
├── imgFlux-upload-api/         # Image upload REST API module
├── imgFlux-download-api/       # Image download and resize module
├── imgFlux-admin-ui/           # Admin UI module
├── imgFlux-mcp-server/         # MCP server for AI agent integration

Quick Start

Docker Compose (Recommended)

git clone 
cd imgFlux
docker-compose up -d -f docker-compose-develop.yml
# Access: http://localhost:8080

Manual Start

mvn clean install
mvn spring-boot:run -pl imgFlux-upload-api
mvn spring-boot:run -pl imgFlux-download-api
mvn spring-boot:run -pl imgFlux-admin-ui
mvn spring-boot:run -pl imgFlux-mcp-server    # MCP server on port 8088

Upload and Generate Thumbnail

Upload an image:

curl http://localhost:8080/api/v1/upload/xxxx/xx.png

The response path is:

xxxx/xx.png

Use the returned path to generate a thumbnail:

Create a thumbnail

Thumbnail URL structure

  http://${yourDomain}/api/v1/thumbnail/resize/${resizeParam}/${url}

Parameters

mode: local or remote
resizeParam: format [width]x[height]q[qulity][ex][trim], example : 300x300q80trim
url: image URL
- local: image key in object storage
- remote: image URL, example: https://brand.github.com/_next/static/media/logo-03.cc5e5332.png

Local mode (image stored in object storage):

path=xxxx/xx.png.webp
curl http://localhost:8080/api/v1/thumbnail/resize/local/240x240q90trim/${path}
// it will return an image with size=240x240, quality=90, convert format from png to webp

Remote mode (image from remote URL):

curl "http://localhost:8080/api/v1/thumbnail/resize/remote/240x240q90trim/https://brand.github.com/_next/static/media/logo-03.cc5e5332.png"
// it will download the image from the remote URL and process it

Alternative endpoint with query parameters:

# Local mode
curl "http://localhost:8080/api/v1/thumbnail/forge/local/folder/image.png?width=240&height=240&quality=90&format=webp"

# Remote mode
curl "http://localhost:8080/api/v1/thumbnail/forge/remote/https://brand.github.com/_next/static/media/logo-03.cc5e5332.png?width=240&height=240&quality=90&format=webp"

Online Demo

Features	url	preview
original	https://brand.github.com/_next/static/media/logo-03.cc5e5332.png
Resize 200x200	http://thumbnail.rnh-inc.com/api/v1/thumbnail/resize/remote/200x200/https://brand.github.com/_next/static/media/logo-03.cc5e5332.png
Trim	http://thumbnail.rnh-inc.com/api/v1/thumbnail/resize/remote/200x200trim/https://brand.github.com/_next/static/media/logo-03.cc5e5332.png
Extent	http://thumbnail.rnh-inc.com/api/v1/thumbnail/resize/remote/200x200ex/https://brand.github.com/_next/static/media/logo-03.cc5e5332.png
Quality 5%	http://thumbnail.rnh-inc.com/api/v1/thumbnail/resize/remote/200x200q5/https://brand.github.com/_next/static/media/logo-03.cc5e5332.png
Convert format	http://thumbnail.rnh-inc.com/api/v1/thumbnail/resize/remote/200x200q80/https://brand.github.com/_next/static/media/logo-03.cc5e5332.png.webp
Text Translation (Korean→English)	not available in Demo
Text Translation (Korean→Chinese)	not available in Demo

Tech Stack

Spring Boot 3.3 - Application framework
Java 17 - Programming language
GraphicsMagick - Image processing engine
MinIO / AWS S3 / Alibaba Cloud OSS - Object storage
JWT - Authentication & authorization
H2 Database - Embedded database
Thymeleaf - Template engine

Development

Any company which access and use this product is welcome to register at the address, only for product promotion purpose.

Welcome everyone’s attention and use, imgFlux will also embrace changes, sustainable development.

Contributing

Contributions are welcome! Please see CONTRIBUTING.md for details.

License

This project is licensed under the GPL License.

Contact

Issue reporting: GitHub Issues