doc-extractor

Crawl documentation sites and pipe them into your Claude Code terminal.

You point it at a docs URL, it extracts every page as markdown, and one click sends the content into your active Claude Code session so you can code against it in your own project.

Install

npx doc-extractor

This starts a local UI on http://localhost:3000 and installs three hooks into ~/.claude/settings.json so Claude Code can receive extractions.

Use

1. Open your project, start Claude Code (claude) in its terminal.
2. Open http://localhost:3000 and paste a docs URL.
3. When the crawl finishes, click Send to Claude on any page (or the whole extraction).
4. Type anything in your Claude terminal — the extracted docs are injected into your prompt and Claude answers grounded in them, with access to your project files.

How it works

┌─────────────┐    writes queue    ┌──────────────┐
│   UI (:3000)│ ─────────────────> │ ~/.claude/   │
└─────────────┘                    │ doc-extractor│
                                   │ -queue/      │
                                   └──────┬───────┘
                                          │ read by UserPromptSubmit hook
                                          ▼
                                   ┌──────────────┐
                                   │ Claude Code  │
                                   │ (your term)  │
                                   └──────────────┘

The crawler uses cheerio + turndown and runs entirely on your machine. Nothing is sent to a third-party service.

Requirements

• Node 18+
• Claude Code installed and on PATH
• bash available (Git Bash on Windows, built-in on macOS/Linux)

Uninstall

npx doc-extractor uninstall

Removes hooks, scripts and the queue directory. Your extractions history is kept in ~/.claude/doc-extractor-history/.

License

MIT