Voice AI Website Analyzer MCP Server

Voice AI Website Analyzer MCP Server

An MCP (Model Context Protocol) server designed for Voice AI in GoHighLevel (GHL) that fetches and analyzes website content to provide business context to AI agents.

Features

• Smart Website Crawling: Fetches homepage + up to 4 important pages (About, Services, Contact, etc.)
• Business Details Extraction: Automatically extracts:
• Business name and description
• Contact information (phone, email, address)
• Business hours
• Services offered
• Text Content Analysis: Provides comprehensive text summaries of each page
• AI-Ready Output: Returns formatted text description perfect for AI agent context

Installation

1. Clone this repository:

git clone <your-repo-url>
cd Voice_MCP

1. Install dependencies:

npm install

1. Build the project:

npm run build

Usage

Running Locally

The MCP server runs on stdio transport:

npm start

Configuring in Claude Desktop or MCP Client

Add to your MCP client configuration (e.g., claude_desktop_config.json):

{
  "mcpServers": {
    "voice-ai-website-analyzer": {
      "command": "node",
      "args": ["d:\\Voice_MCP\\dist\\index.js"]
    }
  }
}

Using the Tool

Once configured, you can use the analyze_website tool:

analyze_website({ url: "https://example.com" })

The tool will:

1. Fetch the homepage
2. Identify and fetch up to 4 important pages (About, Services, Contact, etc.)
3. Extract business details from all pages
4. Return a comprehensive text analysis

Example Output

BUSINESS WEBSITE ANALYSIS
==================================================

Website: https://example.com
Pages Analyzed: 5

BUSINESS DETAILS
--------------------------------------------------
Business Name: Example Business Inc.
Description: We provide excellent services to our customers
Phone: (555) 123-4567
Email: info@example.com
Address: 123 Main Street, City, State 12345
Business Hours: Monday-Friday 9AM-5PM

SERVICES OFFERED
--------------------------------------------------
1. Web Development
2. Mobile App Development
3. Consulting Services
4. Technical Support

PAGE SUMMARIES
--------------------------------------------------

Page 1: Home - Example Business
URL: https://example.com
Content Preview: Welcome to Example Business...

Page 2: About Us
URL: https://example.com/about
Content Preview: Learn more about our company...

Deployment on Vercel

Option 1: Deploy via Vercel CLI

1. Install Vercel CLI:

npm i -g vercel

1. Deploy:

vercel

Option 2: Deploy via GitHub

1. Push your code to GitHub
2. Import the repository in Vercel dashboard
3. Vercel will auto-detect the project and deploy

Important Note About Vercel Deployment

⚠️ MCP servers typically run on stdio transport and are designed to be run locally or on long-running servers. Vercel is optimized for serverless functions with HTTP endpoints.

For production use with GHL Voice AI, consider:

1. Hosting on a VPS (Digital Ocean, AWS EC2, etc.) where the MCP server can run continuously
2. Converting to HTTP API if you need serverless deployment
3. Using Vercel for API endpoints and wrapping the MCP functionality in HTTP handlers

Converting to HTTP API (for Vercel)

If you need to deploy on Vercel, you'll want to create API endpoints instead. Let me know if you need help converting this to an HTTP API format.

Configuration

The server is configured to:

• Fetch maximum of 5 pages total (1 homepage + 4 additional)
• Extract text content (up to 5000 characters per page)
• Identify important pages using keywords: about, services, contact, products, portfolio, team
• Extract common business information patterns

Development

Project Structure

Voice_MCP/
├── src/
│   └── index.ts          # Main MCP server implementation
├── dist/                 # Compiled JavaScript (generated)
├── package.json
├── tsconfig.json
├── vercel.json
└── README.md

Building

npm run build

Testing Locally

After building, run: ``bash node dist/index.js ``

The server will start and wait for MCP protocol messages on stdin.

Integration with GHL Voice AI

When integrated with GHL Voice AI:

1. The AI agent receives the website URL from user input during conversation
2. The agent calls the analyze_website tool with the URL
3. The MCP server fetches and analyzes the website
4. The business context is returned to the AI agent
5. The AI agent uses this context to provide tailored responses about the business

Requirements

• Node.js 18 or higher
• TypeScript 5.x

Dependencies

• @modelcontextprotocol/sdk: MCP protocol implementation
• cheerio: HTML parsing and manipulation
• node-fetch: HTTP requests

License

MIT

Support

For issues or questions, please open an issue in the repository.