webappmcp

webappmcp

Enables AI assistants to interact with web applications through DOM inspection, user interaction simulation, and application state management.

Category
访问服务器

README

WebApp MCP Server

A Model Context Protocol (MCP) server that enables AI assistants to interact with web applications through DOM inspection, user interaction simulation, and application state management.

Features

  • 🔍 DOM Inspection - Query and inspect DOM elements using CSS selectors
  • 🖱️ User Interaction - Simulate clicks, typing, scrolling, and other user actions
  • 📸 Visual Capture - Take screenshots of pages or specific elements
  • 🔧 State Access - Read application state, local storage, and console logs
  • 🚀 Framework Support - Works with React, Vue, Angular, and vanilla JavaScript
  • 🔒 Secure - Built-in authentication and permission controls

Installation

npm install @cgaspard/webappmcp

Quick Start

1. Add to your Express application

import express from 'express';
import { webappMCP } from '@cgaspard/webappmcp';

const app = express();

// Configure the MCP middleware
app.use(webappMCP({
  transport: 'sse',
  wsPort: 4835,
  appPort: 3000,  // Tell middleware what port Express will use
  cors: {
    origin: true,
    credentials: true
  }
}));

app.listen(3000);
// The middleware will display the correct MCP URL when initialized

2. Add client to your frontend

<script src="https://unpkg.com/@cgaspard/webappmcp/dist/browser.min.js"></script>
<script>
  const mcpClient = new WebAppMCP.WebAppMCPClient({
    serverUrl: 'ws://localhost:4835',
    autoConnect: true
  });
  
  mcpClient.connect();
</script>

Or with npm:

import { WebAppMCPClient } from '@cgaspard/webappmcp';

const mcpClient = new WebAppMCPClient({
  serverUrl: 'ws://localhost:4835',
  autoConnect: true
});

mcpClient.connect();

3. Configure your AI assistant

Claude Desktop App

Add using the command line (example for basic todos app):

claude mcp add webapp-sse sse:http://localhost:4834/mcp/sse

For any of the example apps, use the same standardized port:

  • All Examples: http://localhost:4834/mcp/sse

Or manually edit your configuration (example for basic todos app):

{
  "mcpServers": {
    "webapp-sse": {
      "transport": {
        "type": "sse",
        "url": "http://localhost:4834/mcp/sse"
      }
    }
  }
}

Claude Code CLI

Add to your Claude Code configuration (~/.config/claude-code/settings.json):

{
  "mcpServers": {
    "webapp": {
      "transport": {
        "type": "sse", 
        "url": "http://localhost:4834/mcp/sse"
      }
    }
  }
}

Cline (VS Code Extension)

Add to your Cline MCP settings in VS Code:

{
  "webapp": {
    "transport": {
      "type": "sse",
      "url": "http://localhost:4834/mcp/sse"
    }
  }
}

Continue.dev

Add to your Continue configuration (~/.continue/config.json):

{
  "models": [...],
  "mcpServers": {
    "webapp": {
      "transport": {
        "type": "sse",
        "url": "http://localhost:4834/mcp/sse"
      }
    }
  }
}

Zed Editor

Add to your Zed assistant panel settings:

{
  "mcpServers": {
    "webapp": {
      "transport": {
        "type": "sse",
        "url": "http://localhost:4834/mcp/sse"
      }
    }
  }
}

Available Tools

All tools are prefixed with webapp_ to prevent naming conflicts with other MCP servers.

DOM Inspection

  • webapp_dom_query - Find elements using CSS selectors
  • webapp_dom_get_properties - Get element properties and attributes
  • webapp_dom_get_text - Extract text content
  • webapp_dom_get_html - Get HTML structure
  • webapp_dom_manipulate - Modify DOM elements (setAttribute, addClass, etc.)

User Interactions

  • webapp_interaction_click - Click on elements
  • webapp_interaction_type - Type text into inputs
  • webapp_interaction_scroll - Scroll page or elements
  • webapp_interaction_hover - Hover over elements

Visual Capture

  • webapp_capture_screenshot - Take full page screenshots
  • webapp_capture_element_screenshot - Capture specific elements

State Management

  • webapp_state_get_variable - Access JavaScript variables
  • webapp_state_local_storage - Read/write local storage
  • webapp_console_get_logs - Retrieve browser console logs
  • webapp_console_save_to_file - Save browser logs to file

Server-Side Tools

  • webapp_console_get_server_logs - Retrieve Node.js server logs
  • webapp_server_execute_js - Execute JavaScript on the server (sandboxed)
  • webapp_server_get_system_info - Get process and system information
  • webapp_server_get_env - Inspect environment variables (masked)

Diagnostic Tools

  • webapp_list_clients - List connected browser clients
  • webapp_javascript_inject - Execute JavaScript code in the browser
  • webapp_execute_javascript - Execute JavaScript with async support

Terminology Guide

When working with AI assistants using WebApp MCP, use these terms for clarity:

Recommended Terms

  • "the connected web app" - The web page being controlled (preferred)
  • "the browser client" - The frontend/browser instance
  • "the target application" - Formal term for the controlled app
  • "the MCP client" - When discussing the MCP connection

Example Usage

Good:

  • "Click the submit button in the connected web app"
  • "Take a screenshot of the browser client"
  • "Get the current route from the target application"

Avoid:

  • "Click the button" (ambiguous)
  • "Check the page" (which page?)
  • "Get the state" (from where?)

Configuration Options

webappMCP({
  // Transport type: 'sse' (default), 'stdio', 'socket', or 'none'
  transport: 'sse',
  
  // Express app port (defaults to process.env.PORT || 3000)
  appPort: 3000,
  
  // WebSocket port for client connections
  wsPort: 4835,
  
  // MCP SSE endpoint path
  mcpEndpointPath: '/mcp/sse',
  
  // Authentication settings
  authentication: {
    enabled: true,
    token: 'your-secure-token'
  },
  
  // Permission controls
  permissions: {
    read: true,        // Allow DOM reading
    write: true,       // Allow DOM modifications
    screenshot: true,  // Allow screenshots
    state: true        // Allow state access
  },
  
  // CORS settings
  cors: {
    origin: '*',
    credentials: true
  },
  
  // Screenshot storage directory (relative to project root)
  screenshotDir: '.webappmcp/screenshots',
  
  // Debug logging
  debug: false,
  
  // Server-side console log capture
  captureServerLogs: true,     // Enable/disable all server log capture (default: true)
  serverLogLimit: 1000,         // Maximum logs to keep in memory (default: 1000)

  // Winston logger (RECOMMENDED: pass your logger directly)
  winstonLogger: logger,       // Optional Winston logger instance for direct integration

  // Granular log capture configuration
  logCapture: {
    console: false,   // Capture console.log/warn/error/info (disable if using Winston)
    streams: false,   // Capture stdout/stderr streams (disable if using Winston)
    winston: true,    // Capture Winston logs via transport (default: true)
    bunyan: false,    // Capture Bunyan logs (default: true)
    pino: false,      // Capture Pino logs (default: true)
    debug: false,     // Capture debug library logs (default: true)
    log4js: false     // Capture log4js logs (default: true)
  }
});

Server Log Capture

WebApp MCP can capture server-side console logs and logging library output, making them accessible through the MCP tools. This is especially useful for debugging and monitoring.

Features

  • Multi-layer capture: Intercepts logs at library, console, and stream levels
  • Winston support: Direct integration via manual configuration (recommended)
  • Circular buffer: Keeps only the most recent logs (configurable limit)
  • Selective capture: Choose which log sources to capture
  • Performance-friendly: Disable specific interceptors for better performance

Winston Integration (Recommended)

The best way to capture Winston logs is to pass your logger directly:

const winston = require('winston');

// Create Winston logger
const logger = winston.createLogger({
  level: 'info',
  transports: [new winston.transports.Console()]
});

// Pass it to the middleware
const mcpMiddleware = app.use(webappMCP({
  winstonLogger: logger,  // Direct integration (recommended!)
  captureServerLogs: true,
  logCapture: {
    console: false,  // Disable console capture
    winston: true    // Winston capture via winstonLogger param
  }
}));

// Alternative: Attach logger after setup (if created elsewhere)
// mcpMiddleware.attachWinston(logger);

Configuration Examples

// Winston-only capture (recommended for production)
const logger = winston.createLogger({ /* ... */ });

app.use(webappMCP({
  winstonLogger: logger,
  captureServerLogs: true,
  logCapture: {
    console: false,
    streams: false,
    winston: true
  }
}));

// Console only (lightweight, development)
app.use(webappMCP({
  captureServerLogs: true,
  logCapture: {
    console: true,
    streams: false,
    winston: false
  }
}));

// Attach Winston from separate module
const mcpMiddleware = app.use(webappMCP({ captureServerLogs: true }));
const logger = require('./config/logger');
mcpMiddleware.attachWinston(logger);  // Attach after the fact

Examples

Check out the Todos App Example - a fully functional todo application that demonstrates all WebApp MCP features.

Common Use Cases

// Add a new todo
await webapp.interaction.type({ 
  selector: '#new-todo',
  text: 'Buy groceries'
});
await webapp.interaction.click({ selector: '#add-todo' });

// Toggle todo completion
await webapp.interaction.click({ selector: '.todo-checkbox' });

// Filter todos
await webapp.interaction.click({ selector: '[data-filter="active"]' });

// Access application state
const todos = await webapp.state.getVariable({ 
  path: 'window.todosApp.todos' 
});

Security

WebApp MCP Server includes several security features:

  • Authentication - Token-based authentication for MCP connections
  • Rate Limiting - Prevent abuse with configurable rate limits
  • Input Sanitization - All DOM queries are sanitized to prevent XSS
  • Permission Control - Fine-grained control over allowed operations
  • HTTPS Support - Secure WebSocket connections

Framework Integration

React

import { useEffect } from 'react';
import { WebAppMCPClient } from '@cgaspard/webappmcp';

function App() {
  useEffect(() => {
    const client = new WebAppMCPClient({
      serverUrl: 'ws://localhost:4835',
      autoConnect: true
    });
    
    client.connect();
    
    return () => client.disconnect();
  }, []);
  
  return <div>Your app content</div>;
}

Vue

import { WebAppMCPClient } from '@cgaspard/webappmcp';

export default {
  mounted() {
    this.mcpClient = new WebAppMCPClient({
      serverUrl: 'ws://localhost:4835',
      autoConnect: true
    });
    
    this.mcpClient.connect();
  },
  
  beforeUnmount() {
    if (this.mcpClient) {
      this.mcpClient.disconnect();
    }
  }
}

Angular

import { Component, OnInit, OnDestroy } from '@angular/core';
import { WebAppMCPClient } from '@cgaspard/webappmcp';

@Component({
  selector: 'app-root',
  templateUrl: './app.component.html'
})
export class AppComponent implements OnInit, OnDestroy {
  private mcpClient: WebAppMCPClient;
  
  ngOnInit() {
    this.mcpClient = new WebAppMCPClient({
      serverUrl: 'ws://localhost:4835',
      autoConnect: true
    });
    
    this.mcpClient.connect();
  }
  
  ngOnDestroy() {
    if (this.mcpClient) {
      this.mcpClient.disconnect();
    }
  }
}

Development

# Clone the repository
git clone https://github.com/cgaspard/webappmcp.git
cd webappmcp

# Install dependencies
npm install

# Build all packages
npm run build

# Run tests
npm test

# Start development server
npm run dev

VS Code Integration

This project includes full VS Code support for easy development and debugging. See VS_CODE_SETUP.md for details.

Quick start with VS Code:

  1. Open the project in VS Code
  2. Press F5 to launch both the demo app and MCP server
  3. Visit http://localhost:3456 to see the demo

Contributing

We welcome contributions! Please see our Contributing Guide for details.

License

MIT © cgaspard

Support


Built with ❤️ to make AI-powered web automation accessible to everyone.

推荐服务器

Baidu Map

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。

官方
精选
JavaScript
Playwright MCP Server

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。

官方
精选
TypeScript
Magic Component Platform (MCP)

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。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

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。

官方
精选
本地
TypeScript
VeyraX

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。

官方
精选
本地
graphlit-mcp-server

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。

官方
精选
TypeScript
Kagi MCP Server

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。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

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。

官方
精选