MCP Image Compare Server
Enables pixel-perfect image comparison and visual regression testing via tools to compare local images, screenshots, and URLs with detailed diff statistics.
README
MCP Image Compare Server
Tiếng Việt | English
A Model Context Protocol (MCP) server for pixel-perfect image comparison using Pixelmatch from Mapbox. Compare images, capture and compare screenshots, detect visual differences with precision.
Features
- 🎯 Pixel-perfect comparison - Detect even single-pixel differences
- 🎨 Perceptual color difference - YIQ color space for human-like perception
- 🔍 Anti-aliasing detection - Smart detection and exclusion of anti-aliased pixels
- 📊 Detailed statistics - Pixel count, percentage, dimensions
- 🖼️ Visual diff output - Color-coded difference visualization
- 🌐 Web screenshot support - Compare with live websites
- 📁 Multiple formats - PNG and JPEG support
- ⚡ Fast & lightweight - ~60-80KB package size
Quick Start
Installation
# Install globally
npm install -g mcp-image-compare-server
# Install Chromium for screenshots
npx playwright install chromium
Configuration
For Claude Desktop
Add to your Claude Desktop config file:
Windows: %APPDATA%\Claude\claude_desktop_config.json
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Linux: ~/.config/Claude/claude_desktop_config.json
{
"mcpServers": {
"image-compare": {
"command": "npx",
"args": ["-y", "mcp-image-compare-server"]
}
}
}
Restart Claude Desktop after configuration.
For Cursor
Add to your Cursor MCP settings file:
Windows: %APPDATA%\Cursor\User\globalStorage\mcp.json
macOS: ~/Library/Application Support/Cursor/User/globalStorage/mcp.json
Linux: ~/.config/Cursor/User/globalStorage/mcp.json
{
"mcpServers": {
"image-compare": {
"command": "npx",
"args": ["-y", "mcp-image-compare-server"]
}
}
}
Or through Cursor settings:
- Open Cursor Settings (
Ctrl+,orCmd+,) - Search for "MCP"
- Click "Edit in settings.json"
- Add the configuration above
Restart Cursor after configuration.
Usage
Ask in Claude Desktop or Cursor:
Compare image1.png and image2.png
That's it! 🎉
Tools
1. compare_images
Compare two local image files.
Parameters:
image1_path(string, required) - Path to first imageimage2_path(string, required) - Path to second imagediff_output_path(string, optional) - Where to save diff imagethreshold(number, optional) - Comparison threshold (0-1), default 0.1includeAA(boolean, optional) - Include anti-aliased pixels, default false
Example:
{
"image1_path": "./screenshots/before.png",
"image2_path": "./screenshots/after.png",
"diff_output_path": "./diff-result.png",
"threshold": 0.1
}
2. compare_image_with_url
Compare a local image with a screenshot from a URL.
Parameters:
image_path(string, required) - Path to local imageurl(string, required) - URL to capture and comparediff_output_path(string, optional) - Where to save diff imagethreshold(number, optional) - Comparison threshold (0-1), default 0.1includeAA(boolean, optional) - Include anti-aliased pixels, default false
Example:
{
"image_path": "./design-mockup.png",
"url": "https://example.com",
"threshold": 0.15
}
3. compare_urls
Compare screenshots from two different URLs.
Parameters:
url1(string, required) - First URL to captureurl2(string, required) - Second URL to capturediff_output_path(string, optional) - Where to save diff imagethreshold(number, optional) - Comparison threshold (0-1), default 0.1includeAA(boolean, optional) - Include anti-aliased pixels, default false
Example:
{
"url1": "https://staging.example.com",
"url2": "https://production.example.com"
}
Output Format
All tools return JSON with detailed statistics:
{
"success": true,
"diffPixels": 1234,
"totalPixels": 921600,
"percentDiff": 0.13,
"width": 1280,
"height": 720,
"diffImagePath": "/path/to/diff.png",
"message": "Comparison completed: 1234 pixels differ (0.13%)"
}
Fields:
success- Operation status (true/false)diffPixels- Number of different pixelstotalPixels- Total pixels in imagepercentDiff- Percentage differencewidth- Image width in pixelsheight- Image height in pixelsdiffImagePath- Path to generated diff imagemessage- Human-readable result messageerror- Error message (if failed)
Use Cases
Visual Regression Testing
Compare UI before and after updates to catch unintended changes.
I have two screenshots:
- before-update.png
- after-update.png
Compare them and tell me what changed.
Cross-browser Testing
Compare rendering across different browsers to ensure consistency.
Design Review
Compare design mockups with actual implementation.
Compare design-mockup.png with https://myapp.com/landing
A/B Testing
Compare different versions of the same page.
Compare:
- https://myapp.com/variant-a
- https://myapp.com/variant-b
Mobile vs Desktop
Compare responsive layouts across different viewports.
Configuration
Threshold Values
The threshold parameter (0-1) controls comparison sensitivity:
- 0.0 - Very sensitive, detects smallest differences
- 0.1 - Default, good balance for most cases
- 0.3 - Less sensitive, ignores minor differences
- 1.0 - Least sensitive, only major differences
Anti-aliasing
When includeAA = false (default), the server automatically detects and ignores anti-aliased pixels, reducing false positives from different rendering engines.
Diff Image Colors
- Red: Different pixels
- Yellow: Anti-aliased pixels (if includeAA = true)
- Faded: Matching pixels (reduced opacity)
Installation Methods
Method 1: NPM Global (Recommended)
npm install -g mcp-image-compare-server
npx playwright install chromium
Configure Claude Desktop:
{
"command": "npx",
"args": ["-y", "mcp-image-compare-server"]
}
Method 2: NPX (No Installation)
Configure Claude Desktop:
{
"command": "npx",
"args": ["-y", "mcp-image-compare-server"]
}
First run will download the package automatically.
Method 3: From Source
git clone https://github.com/leky90/mcp-image-compare-server.git
cd mcp-image-compare-server
npm install
npx playwright install chromium
npm run build
Configure Claude Desktop or Cursor:
{
"command": "node",
"args": ["/absolute/path/to/mcp-image-compare-server/dist/index.js"]
}
Replace /absolute/path/to/mcp-image-compare-server with your actual path.
System Requirements
- Node.js: 18.0.0 or higher
- npm: 8.0.0 or higher
- OS: Windows 10+, macOS 10.15+, Linux (Ubuntu 20.04+)
- Disk: ~500MB (includes Chromium browser)
- RAM: 2GB minimum
Troubleshooting
Server doesn't connect
- Verify config path is correct
- Ensure you ran
npm run build - Restart Claude Desktop or Cursor completely
- Check the MCP logs for errors
"Browser not found" error
npx playwright install chromium
Screenshot timeout
For slow websites, the default 30-second timeout will handle most cases. If issues persist, please report.
Out of memory
For very large images:
- Resize images before comparison
- Use lower resolution images
- Increase Node.js memory:
node --max-old-space-size=4096 dist/index.js
Development
Building from Source
npm install
npm run build
Running Locally
npm start
# or
npm run dev
Testing
See CONTRIBUTING.md for development guidelines.
Tech Stack
- Framework: Model Context Protocol SDK
- Image Comparison: Pixelmatch
- Screenshots: Playwright
- Image Processing: Sharp, pngjs
- Language: TypeScript
Contributing
Contributions are welcome! Please read CONTRIBUTING.md for guidelines.
Changelog
See CHANGELOG.md for version history.
License
MIT License - see LICENSE file for details.
Acknowledgments
Built with:
- Pixelmatch by Mapbox
- Model Context Protocol
- Playwright
- Sharp
Support
- 🐛 Bug Reports: GitHub Issues
- 💬 Questions: GitHub Discussions
- 📖 Documentation: This README
Made with ❤️ using TypeScript and MCP
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。