Web Scraper MCP Server

An MCP (Model Context Protocol) server that can scrape web pages and extract content using CSS selectors. Built with deno-dom for fast HTML parsing.

Why

Most LLM clients already have some HTTP fetching capabilities, but fetching a page directly often returns a lot of unnecessary content. This not only confuses the LLM, but also quickly fills up the context window.

That's where this MCP comes in—it enables targeted scraping using CSS selectors, so you only extract the content you actually need.

Example from Zed

See ZedExample.md for a real-world usage example.

Features

• 🌐 Fetch any publicly accessible web page by URL
• 🔍 Parse HTML content using the fast deno-dom library
• 📋 Extract text content using standard CSS selectors
• 🎯 Support for complex selectors (classes, IDs, attributes, pseudo-selectors)
• ⚡ Built-in error handling for network issues and parsing failures
• 🛡️ Safe execution with minimal required permissions

Prerequisites

• Deno installed on your system
• Network access for fetching web pages

Running the Server

deno run --allow-net jsr:@sigma/scrap-mcp

You can also run this with Bun and Node.js using bunx and npx respectively:

bunx rjsr @sigma/scrap-mcp

npx rjsr @sigma/scrap-mcp

MCP Tool Reference

scrape_page

The main tool for scraping web pages and extracting content.

Parameters:

• url (string, required): The URL of the page to scrape
• query_selector (string, required): CSS selector to query elements

Return Format:

Found X elements matching selector "SELECTOR" on URL:

Element 1: TEXT_CONTENT

Element 2: TEXT_CONTENT
...

Usage Examples

Basic Selectors

1. 1.Extract all headings:

```json

{ "url": "https://example.com", "query_selector": "h1, h2, h3" }

```

1. 1.Extract all paragraphs:

```json

{ "url": "https://example.com", "query_selector": "p" }

```

1. 1.Extract content from specific classes:

```json

{ "url": "https://news.ycombinator.com", "query_selector": ".titleline > a" }

```

1. 1.Extract all links:

```json

{ "url": "https://example.com", "query_selector": "a" }

```

Advanced Selectors

1. 1.Extract navigation items:

```json

{ "url": "https://deno.land", "query_selector": "nav a" }

```

1. 1.Extract elements with specific attributes:

```json

{ "url": "https://example.com", "query_selector": "a[href^='https://']" }

```

1. 1.Extract form inputs:

```json

{ "url": "https://example.com", "query_selector": "input[type='text'], input[type='email']" }

```

CSS Selector Reference

Basic Selectors

• h1 - All H1 headings
• .className - All elements with class "className"
• #elementId - Element with ID "elementId"
• * - All elements

Combinators