home / mcp / fetcher mcp server

Fetcher MCP Server

Fetch web page content using a Playwright headless browser with intelligent content extraction and Markdown/HTML output.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

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

Fetcher MCP is a Node-based server that uses a headless Playwright browser to fetch and extract content from web pages. It supports JavaScript rendering, intelligent content extraction, and multiple output formats, making it ideal for batch scraping and content analysis while minimizing bandwidth usage.

How to use

You run Fetcher MCP from the command line and connect to it with an MCP client. It exposes a streamable HTTP endpoint and an SSE endpoint for different MCP client workflows. Use the simple npx command to start the server, then choose the transport you prefer and connect accordingly.

How to install

Prerequisites: you need Node.js installed on your system. You will also install a Playwright browser binary for Chromium so the headless browser can render pages.

Step-by-step commands to install and run Fetcher MCP locally:

# Install dependencies for the project (if applicable in your setup)
npm install

# Install Playwright browsers needed for Chromium
npm run install-browser

# Start the MCP server using npx
npx -y fetcher-mcp

Configuration, deployment, and tips

Configure how you deploy and access Fetcher MCP in different environments. The server can be run with an HTTP transport to provide both a Streamable HTTP endpoint and an SSE endpoint for legacy MCP clients.

Debug mode lets you watch the browser during operation for troubleshooting. You can enable it for a specific operation or run the server with the debug flag.

You can run Fetcher MCP in Docker for production or testing. The container exposes port 3000 and can be extended with a health check to ensure the service stays healthy.

HTTP and SSE transport and endpoints

To enable the HTTP transport that exposes the Streamable HTTP endpoint and the SSE endpoint, start the server with the HTTP flag. After startup, you will access the endpoints at the following paths on your host.

- /mcp — Streamable HTTP endpoint (modern MCP protocol) - /sse — SSE endpoint (legacy MCP protocol)

Browser and content extraction features

Fetcher MCP uses a headless Playwright browser to render pages, then applies a built-in Readability-like algorithm to extract the main content. It can output either HTML or Markdown, and can optimize resource usage by blocking images, styles, fonts, and media during fetches.

Available tools

fetch_url

Retrieve a single web page content by URL using a Playwright headless browser, with optional content extraction and Markdown/HTML output.

fetch_urls

Fetch content from multiple URLs in parallel, returning combined results with clear separation between pages.

browser_install

Install the Playwright Chromium browser binary automatically when not present, with option to install required system dependencies.