sargel
Lets AI agents visually inspect web elements, test CSS edits in real-time, and iterate until pixel-perfect, functioning like browser DevTools for debugging UI issues.
README
SARGEL - Style And Rendering Graphical Element Lens
Visual debugging for pixel-perfect web development.
Whether you're implementing designs or debugging CSS issues, this tool gives AI agents the ability to see and fix UI problems visually - just like using browser DevTools. Inspect elements, test CSS changes instantly, and iterate until perfect.
<img src="docs/images/hero-screenshot.png" width="800" alt="AI agent using SARGEL to visually inspect and debug a web element with highlighted overlays and detailed CSS analysis">
The Problem
Getting CSS exactly right is hard. Designs don't match implementation, layouts break across browsers, spacing is off by a few pixels. The usual fix: back-and-forth screenshots, trial and error, and "close enough" compromises.
The solution: AI agents that can see your UI, understand what's wrong visually, test fixes instantly, and iterate until pixel-perfect.
How It Works
<img src="docs/images/css-workflow.gif" width="700" alt="Visual demonstration of the inspect-edit-verify-iterate workflow">
- Inspect - AI sees current state: styles, layout, visual issues
- Test - Apply CSS edits and see results immediately
- Verify - Get screenshot with changes applied
- Iterate - Refine until perfect
- Apply - Copy working CSS to your code
The key: Test changes visually before touching your source code.
Example: Fixing Button Spacing
Problem: "The submit button is too close to the form fields"
// Step 1: Inspect current state
{
"tool": "inspect_element",
"arguments": {
"css_selector": ".submit-btn",
"url": "https://myapp.com/form"
}
}
// Returns: Screenshot + computed styles showing margin-top: 4px
// Step 2: Test a fix
{
"tool": "inspect_element",
"arguments": {
"css_selector": ".submit-btn",
"url": "https://myapp.com/form",
"css_edits": {
"margin-top": "16px"
}
}
}
// Returns: Updated screenshot showing improved spacing
// Step 3: Perfect it
{
"css_edits": {
"margin-top": "20px"
}
}
// Returns: Perfect spacing
// Result: Add margin-top: 20px to your CSS
<img src="docs/images/css-edits-before.png" width="400" alt="Before: Button too close to form fields"> <img src="docs/images/css-edits-after.png" width="400" alt="After: Perfect button spacing">
Before and after: Visual iteration until pixel-perfect
API Reference
inspect_element
| Parameter | Type | Description |
|---|---|---|
css_selector |
string | CSS selector for target element(s) |
url |
string | Webpage URL to inspect |
css_edits |
object | Test CSS changes before applying to source |
property_groups |
array | Focus on: "layout", "colors", "typography", "positioning" |
limit |
number | Max elements when selector matches multiple (default: 10) |
Response Format
{
screenshot: string; // Base64 PNG with visual highlights
computed_styles: object; // Actual CSS property values
cascade_rules: array; // CSS rules in cascade order
box_model: object; // Margin, padding, border, content dimensions
applied_edits?: object; // CSS edits that were tested (if any)
}
Technical Details
Architecture: MCP (Model Context Protocol) server using Chrome DevTools Protocol for DOM inspection and screenshot capture.
Requirements:
- Node.js 18+
- Chrome/Chromium browser
- AI agent with MCP support (Claude Desktop, Continue, etc.)
License
MIT License - see LICENSE for details.
Made for anyone who believes getting CSS right shouldn't be guesswork
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。