deepwiki-mcp

Why we included this project

Cursor and Claude Desktop users who keep asking the model to explain how a library works under the hood will get real mileage from this MCP server. Point it at a DeepWiki URL, or just give it an owner/repo pair, and it crawls the relevant pages, strips out navigation and ads, and hands the model clean Markdown it can actually read. That turns prompts like "how do I use gpt-image-1 with the Vercel AI SDK" into answers drawn from the project's own docs instead of whatever the model half-remembers from training. You can ask for one combined document or a set of pages, and tune crawl depth and concurrency. Fair warning: DeepWiki has since blocked scraping, so the maintainer says it may not work right now and points people to the official DeepWiki MCP server; the codebase is still a tidy example of how to structure an MCP tool if you want to build your own.