defuddle

To extract clean markdown from a webpage, run the Defuddle CLI with the parse command and the --md flag, which strips away navigation, ads, and repetitive clutter to isolate the main readable text.

The best way to reduce tokens for AI summarization is to parse the web page into clean markdown first, removing irrelevant elements like navigation and ads before submitting the text.

Yes, you can extract targeted metadata fields such as the title, description, or domain from a web page when you only need key page attributes instead of the full article text.

This webpage parsing approach does not work with all URL types; it explicitly avoids URLs ending in .md, where WebFetch should be used instead, but handles standard articles and online documentation.

You should not use webpage parsing for markdown extraction when dealing with URLs ending in .md, because WebFetch should be used instead to retrieve the raw markdown content directly.