doc2mcp

doc2mcp turns any documentation site into a hosted MCP (Model Context Protocol) server that Cursor, Claude Desktop, Windsurf, and other AI clients can read.

How it works

1. Open the chat, flip the doc2mcp toggle, paste a docs URL.
2. doc2mcp crawls the site (Mintlify, Docusaurus, GitBook, OpenAPI, GitHub raw markdown — all supported), preserves code blocks, and indexes everything.
3. You get a remote URL + Bearer token. Paste the JSON into Cursor and reload. Done.

No local install, no npx, no API keys for the original docs vendor.

{  "mcpServers": {    "stripe": {      "url": "https://doc2mcp.dev/api/mcp/<projectId>/mcp",      "headers": {        "Authorization": "Bearer <project-token>"      }    }  }}

What your AI can do

The generated MCP exposes these tools to Cursor:

• list_documentation_pages — every crawled page
• get_documentation_page — full text of a page
• search_documentation — keyword search
• read_full_documentation — all pages in one markdown blob
• ask_documentation — natural-language Q&A with citations

Next

• Getting started
• API keys explained
• MCP setup in Cursor
• Pipeline details