The unofficial DeepWiki reader that can no longer read

What it does It was built to plug DeepWiki documentation into MCP-compatible editors like Cursor. You’d point it at a deepwiki.com URL and it would crawl the site, strip headers, footers, and ads, rewrite links, and hand back clean Markdown—either as one concatenated document or split by page. The README also mentions an NLP feature for searching by library name, though it offers no detail on how it works.

The interesting bit The project treated DeepWiki as a read-only knowledge base for AI coding assistants, essentially turning a documentation website into a structured context window. That worked until DeepWiki explicitly closed the door on scrapers.

Key highlights

• Returns aggregated Markdown or a structured list of pages from crawled docs
• Strips headers, footers, navigation, scripts, and ads, then rewrites links for clean Markdown
• Adjustable crawl depth and concurrency for large repositories
• Exposes a single deepwiki_fetch tool with aggregate or per-page output modes

Caveats

• Currently broken: DeepWiki has blocked scraping, so the server cannot fetch content; the README points users to the official DeepWiki MCP server instead
• Docker deployment is explicitly marked as untested
• The NLP search feature is mentioned in the README but not meaningfully explained

Verdict Worth watching if DeepWiki opens up again or if you want a template for sanitizing web docs into MCP context. Everyone else should follow the README’s advice and use the official server for now.