Summary
The only decision this server asks you to make is how much of the page you want, and it keeps that decision cheap to revise.
Each tool takes url and nothing else, so discovering that a page renders itself client-side costs one more call to get_rendered_html rather than a configuration change. The vendor's own steer is that get_markdown is the default for readable extraction and get_markdown_summary the choice when only the article body matters, with get_raw_text reserved for content where rendering would be wasted work.
What it is
A web-fetching server that returns one URL in a chosen representation, from unrendered source through to a de-chromed Markdown extract.
What you get
- get_raw_text retrieves the source at a url without browser rendering, which is the direct path for structured formats such as JSON, XML, CSV, TSV and plain-text files.
- get_rendered_html runs the page through a headless browser first, so JavaScript-generated content and single-page applications come back complete.
- get_markdown converts a page to Markdown while preserving structural elements including tables and definition lists.
- get_markdown_summary extracts the main content area and drops navigation menus, headers, footers and other peripheral content.
Requirements
Nothing — no account, no key.
Setup effort
One command — npx -y mcp-server-fetch-typescript