Fetch MCP

Name: Fetch MCP
Author: jae-jae

@jae-jae

代码仓库主页

关于 Fetch MCP

MCP server for fetch web page content using Playwright headless browser.

#fetch #web-scraping #playwright

基本信息

分类

浏览器自动化

许可证

MIT

运行时

node

传输方式

stdio

发布者

jae-jae

配置

使用下面的配置,将此服务器添加到你的 MCP 客户端。

{
  "mcpServers": {
    "fetch": {
      "command": "npx",
      "args": [
        "-y",
        "fetcher-mcp"
      ]
    }
  }
}

工具

未检测到工具

工具是从 README 中自动提取的。维护者可以在 ## Tools 标题下列出工具,即可填充这部分内容。

概览

What is Fetch MCP?

Fetch MCP is an MCP server that uses Playwright headless browser to fetch web page content. It handles dynamic JavaScript-rendered pages and intelligently extracts main content, supporting both HTML and Markdown output. Designed for developers integrating web content into MCP-compatible applications like Claude Desktop.

How to use Fetch MCP?

Run directly via npx -y fetcher-mcp after installing the Chromium browser with npx playwright install chromium. Configure the server in your MCP client (e.g., Claude Desktop) by adding its command to the claude_desktop_config.json. Optionally start with --transport=http for HTTP/SSE endpoints, or use Docker deployments.

Key features of Fetch MCP

JavaScript execution via Playwright for dynamic content
Intelligent main content extraction (Readability algorithm)
Supports both Markdown and HTML output formats
Parallel fetching of multiple URLs (fetch_urls tool)
Blocks unnecessary resources (images, styles, fonts, media)
Configurable parameters (timeout, extraction, output format)

Use cases of Fetch MCP

Fetching modern web applications that rely on JavaScript rendering
Batch collecting content from multiple URLs simultaneously
Handling sites with anti-crawler mechanisms (CAPTCHA, redirects)
Preserving original HTML structure when extraction is insufficient
Using debug mode to manually authenticate via cookies

FAQ from Fetch MCP

How is Fetch MCP different from traditional web scrapers?

It uses Playwright to execute JavaScript, enabling it to handle dynamic web content and modern single-page applications that traditional scrapers cannot.

What are the runtime requirements?

Requires Node.js and a Chromium browser installed via Playwright (npx playwright install chromium). Docker images are also available.

How do I get HTML instead of Markdown?

Set the returnHtml parameter to true in the fetch_url or fetch_urls call, or include a prompt like "Please return the content in HTML format."

How to handle anti-crawler verification pages?

Use the waitForNavigation parameter (or prompt "Please wait for the page to fully load") to allow time for CAPTCHAs or redirects to complete. Increase timeout if needed.

What transport protocols are supported?

Supports both Streamable HTTP (endpoint /mcp) and SSE (endpoint /sse) when started with --transport=http. Also supports standard MCP stdio transport.