Summary
A docs site you open every day stops being a URL you paste and becomes something the model can pick off the tool list.
The configuration file is the real product: the wired-in set — React, Next.js, Tailwind CSS, MDN, Cursor and a group of OpenAPI and Swagger specifications — shows what the file can hold rather than fixing what it holds. Every tool is a single fetch and convert, so there is no crawl depth, no link following and no cache: one call brings back one document.
What it is
A fetch-and-convert server that renders a web page as markdown, with a configured list of sites that each become a tool of their own.
What you get
- fetch_website, which converts any URL you hand it into markdown.
- Documentation sites already wired in as their own tools: fetch_react for react.dev, fetch_nextjs for nextjs.org, fetch_tailwind_css for tailwindcss.com, fetch_mdn_web_docs for developer.mozilla.org and fetch_cursor_docs for docs.cursor.com.
- API specifications on the same footing: fetch_github_api for the GitHub REST API OpenAPI description, fetch_openapi_spec for the OpenAPI 3.1.0 specification, and fetch_petstore_openapi, fetch_petstore_swagger and fetch_swagger_ui_demo for the Swagger Petstore in OpenAPI 3.0, Swagger 2.0 JSON and Swagger 2.0 YAML.
- list_configured_websites, which reports which sites the current configuration has turned into tools.
Requirements
WEBSITES_CONFIG_PATH, pointing at the configuration file that decides which sites get their own tool, and network access from the host to the sites being fetched.
Setup effort
One command — npx -y website-to-markdown-mcp