What it does
MarkGrab takes a URL and returns its content as markdown intended for LLM input. For HTML it uses BeautifulSoup with content density filtering to remove navigation, sidebars and ads. It also extracts YouTube transcripts with timestamps, text with page structure from PDFs, and paragraphs and headings from DOCX files. Extraction starts with a lightweight httpx request. If an HTML page yields fewer than 50 words, it retries with Playwright to handle JavaScript-rendered content.
It suits pipelines that need readable text from mixed web sources. The README names newswatch, an RSS news monitoring pipeline, and watchdeck, a web page monitoring tool with visual diffs, as projects that use it. The library is async-first, built on the httpx and Playwright async APIs.
It is used as a Python library through a Python API or a CLI, and returns an ExtractResult object. Optional extras are installed for specific content types. Playwright is needed for the JavaScript fallback. The publisher describes it as part of the QuartzUnit collection of Python libraries.
Limitations and requirements
The publisher states that MarkGrab does not check or enforce robots.txt, so users must check it and the terms of service of each site themselves. It has no built-in rate limiting, so users must implement their own to avoid overloading servers. YouTube transcripts come from the third-party youtube-transcript-api library, which uses an unofficial caption API that may not comply with YouTube's Terms of Service. The optional stealth=True mode alters browser fingerprinting signals to reduce bot detection, and users are responsible for lawful use. Users are responsible for compliance with applicable laws. The software is provided as is, without warranty, under the MIT license.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 13 Nov 2026 · its page on the source
MarkGrab
Universal web content extraction — any URL to LLM-ready markdown. HTML, YouTube, PDF, DOCX.
v0.1.2As shown on registry.modelcontextprotocol.io when we last read it