MCP Server - Playwright HTML Render

This project exposes HTML analysis and correction capabilities via the MCP (Model Context Protocol).

🚀 Installation

npm install

Make sure you have configured your Mistral API key in a .env file:

MISTRAL_API_KEY=your-api-key

📖 Usage

Option 1: Using with npx (No Installation Required)

You can use this MCP server directly with npx without cloning the repository:

npx github:gplanchat/server-playwright

The MCP server communicates via stdio (standard input/output), allowing MCP clients to connect to it.

Note: Make sure you have the MISTRAL_API_KEY environment variable set, or create a .env file in your current directory.

Option 2: Local Installation

If you've cloned the repository:

npm install
npm run mcp

Configuration for Cursor

To use this MCP server with Cursor, add the following configuration to your Cursor settings:

1. Open Cursor settings

2. Search for "MCP" or "Model Context Protocol"

3. Add the following configuration:

Using npx (Recommended):

{
  "mcpServers": {
    "playwright-html-render": {
      "command": "npx",
      "args": ["-y", "github:gplanchat/server-playwright"],
      "env": {
        "MISTRAL_API_KEY": "your-api-key"
      }
    }
  }
}

Using local installation:

{
  "mcpServers": {
    "playwright-html-render": {
      "command": "node",
      "args": ["/path/to/server-playwright/src/mcp-server.js"],
      "env": {
        "MISTRAL_API_KEY": "your-api-key"
      }
    }
  }
}

Note:

• For npx usage, the -y flag automatically accepts the package installation prompt

• For local installation, replace /path/to/server-playwright with the absolute path to your project

Configuration for Claude Desktop

To use with Claude Desktop, modify the MCP configuration file (usually ~/Library/Application Support/Claude/claude_desktop_config.json on macOS or %APPDATA%\Claude\claude_desktop_config.json on Windows):

Using npx (Recommended):

{
  "mcpServers": {
    "playwright-html-render": {
      "command": "npx",
      "args": ["-y", "github:gplanchat/server-playwright"],
      "env": {
        "MISTRAL_API_KEY": "your-api-key"
      }
    }
  }
}

Using local installation:

{
  "mcpServers": {
    "playwright-html-render": {
      "command": "node",
      "args": ["/path/to/server-playwright/src/mcp-server.js"],
      "env": {
        "MISTRAL_API_KEY": "your-api-key"
      }
    }
  }
}

🛠️ Available Tools

The MCP server exposes the following tools:

1. capture_html_render

Captures the render of an HTML page (screenshot, DOM, metadata).

Parameters:

• htmlPath (required): Path to HTML file or URL

• viewport (optional): Viewport size { width: 1920, height: 1080 }

• fullPage (optional): Capture the entire page (default: false)

Usage example:

{
  "name": "capture_html_render",
  "arguments": {
    "htmlPath": "https://example.com",
    "viewport": { "width": 1920, "height": 1080 }
  }
}

2. analyze_html_render

Analyzes an HTML render with AI vision to detect issues.

Parameters:

• screenshotPath (required): Path to screenshot PNG file

• domPath (required): Path to DOM HTML file

• metadataPath (required): Path to metadata JSON file

• criteria (optional): Analysis criteria (default: "default")

Usage example:

{
  "name": "analyze_html_render",
  "arguments": {
    "screenshotPath": "output/example-1234567890.png",
    "domPath": "output/example-1234567890.html",
    "metadataPath": "output/example-1234567890.json",
    "criteria": "default"
  }
}

3. fix_html_render

Fixes an HTML file based on an analysis.

Parameters:

• originalHtmlPath (required): Path to original HTML file

• analysisPath (required): Path to analysis JSON file

• autoApply (optional): Automatically apply corrections (default: false)

Usage example:

{
  "name": "fix_html_render",
  "arguments": {
    "originalHtmlPath": "example.html",
    "analysisPath": "output/analysis-1234567890.json",
    "autoApply": false
  }
}

4. test_and_fix_html

Tests and fixes a complete HTML render (capture + analysis + correction).

Parameters:

• htmlPath (required): Path to HTML file or URL

• viewport (optional): Viewport size { width: 1920, height: 1080 }

• autoFix (optional): Automatically generate corrections (default: false)

• criteria (optional): Analysis criteria (default: "default")

Usage example:

{
  "name": "test_and_fix_html",
  "arguments": {
    "htmlPath": "https://example.com",
    "viewport": { "width": 1920, "height": 1080 },
    "autoFix": true,
    "criteria": "default"
  }
}

5. verify_html_render

Verifies that an HTML render matches an expected detailed description or CSS specifications.

Parameters:

• htmlPath (required): Path to HTML file or URL to verify

• expectedDescription (required): Detailed description of expected render or CSS specifications

• viewport (optional): Viewport size { width: 1920, height: 1080 }

• fullPage (optional): Capture the entire page (default: false)

• verificationType (optional): Verification type - "auto" (automatic detection), "description" (textual description), "css" (CSS specifications) (default: "auto")

Example with textual description:

{
  "name": "verify_html_render",
  "arguments": {
    "htmlPath": "https://example.com",
    "expectedDescription": "The page must contain a header with a logo on the left, a centered navigation menu with 5 links (Home, About, Services, Portfolio, Contact), and a 'Contact Us' button on the right. The header must have a white background and a height of 80px.",
    "verificationType": "description"
  }
}

Example with CSS specifications:

{
  "name": "verify_html_render",
  "arguments": {
    "htmlPath": "example.html",
    "expectedDescription": ".header { background-color: #ffffff; height: 80px; display: flex; align-items: center; justify-content: space-between; } .logo { width: 150px; height: 50px; } .nav-menu { display: flex; gap: 20px; } .nav-menu a { color: #333; text-decoration: none; }",
    "verificationType": "css"
  }
}

Response:

{
  "success": true,
  "conform": true,
  "score": 95,
  "discrepanciesCount": 1,
  "discrepancies": [
    {
      "severity": "minor",
      "property": "gap",
      "expected": "20px",
      "actual": "15px",
      "selector": ".nav-menu",
      "description": "The spacing between menu items is 15px instead of 20px",
      "suggested_fix": "Add gap: 20px; to .nav-menu"
    }
  ],
  "verificationPath": "output/verification-1234567890.md",
  "resultPath": "output/verification-1234567890.json"
}

🧪 Testing with MCP Inspector

You can test the MCP server with the official inspector:

npx @modelcontextprotocol/inspector node src/mcp-server.js

This will open a web interface to test the available tools.

📝 Notes

• The MCP server uses stdio for communication, so it must be launched by an MCP client

• Generated files are saved in the output/ folder

• Make sure Playwright is installed: npx playwright install chromium

• The Mistral API key is required for analysis and correction tools

📚 Additional Documentation

• NPX_USAGE.md - Guide for using the server with npx without cloning

• QUICKSTART.md - Quick start guide

• MISTRAL_SETUP.md - Mistral AI configuration guide

• INTEGRATION.md - AI agent integration guide

🔧 Troubleshooting

Server won't start

Check that:

• Node.js 18+ is installed

• Dependencies are installed (npm install)

• The Mistral API key is configured

Capture errors

Make sure Playwright is installed:

npx playwright install chromium

Analysis errors

Verify that your Mistral API key is valid and you have credits available.