Summary

The design is sound; the source closed the door.

A single call that turns a wiki into clean Markdown at the depth you choose is exactly the right shape for feeding a model. It stopped working when DeepWiki blocked scraping, and the repository now recommends the official server rather than pretending otherwise.

What it is

An unofficial single-tool fetcher for deepwiki.com. Give it a DeepWiki URL, or just a user/repo shortform, and it crawls the relevant pages, strips the navigation and scripts, rewrites the links for Markdown, and returns either one aggregated document or a list of pages.

What you get

A DeepWiki repository page returned as Markdown, aggregated into one document or split into pages with their paths
Crawl depth set on the call, so a single page and a whole wiki are the same tool
Headers, footers, navigation, scripts and ads stripped before conversion, and links rewritten to work in Markdown
Only deepwiki.com URLs accepted, so the crawler cannot be pointed anywhere else
Page count, byte total and elapsed time returned alongside the content

Requirements

Node, run over stdio from npm as mcp-deepwiki, and network access to deepwiki.com. That last part is the catch: DeepWiki has cut off scraping, so this server does not currently return content — the project itself says so and points to the official DeepWiki server maintained by Devin instead.

Setup effort

One command — npx -y mcp-deepwiki@latest