fetcher-mcp

What it does

This MCP server exposes web content fetching via Playwright, a browser automation library. Unlike standard HTTP clients that retrieve raw HTML, this server renders pages in a real browser engine, executing JavaScript and handling dynamic content. It makes this capability available to Claude and other MCP clients through tools for fetching and rendering web pages, capturing the rendered DOM, or retrieving fully-processed HTML after the page stabilizes.

Who it's for

Developers building Claude integrations that need to access JavaScript-rendered content. This is particularly useful for agentic systems interacting with modern single-page applications or websites that serve content only after client-side rendering.

Common use cases

• Fetch and extract content from JavaScript-heavy sites that reject static HTTP requests
• Retrieve fully-rendered HTML after a page's DOM stabilizes and async data loads
• Automate interactions with form fields and dynamic UI elements via Claude
• Capture page state or screenshots after complex client-side behavior completes

Setup pitfalls

• Playwright requires browser binaries (Chromium, Firefox, or WebKit). Run playwright install after setting up the server; initial download is ~300MB.
• Browser processes are memory-intensive. Each fetch may spawn or reuse a browser instance; monitor resource usage in containerized or resource-constrained environments.
• Pages with infinite scroll, slow loads, or rate-limiting may timeout. Configure page load timeouts appropriately and respect target sites' policies.
• Network requests from the browser originate from the server's network context, not your host. Be aware of IP-based restrictions or rate limits the page may enforce.