docrawl is a lightweight Node.js CLI for crawling documentation sites and converting them into Markdown with defuddle.

It is built for static and server-rendered docs sites such as Docusaurus, VitePress, MkDocs, GitBook exports, and Obsidian Publish. It does not run a browser and does not execute page JavaScript.

Why

docrawl is useful when you want to:

• turn docs sites into Markdown for LLM context
• build local knowledge bases
• feed content into RAG pipelines
• archive clean docs content without a browser dependency

Requirements

• Node.js >= 20

Install

Run without installing:

Install globally:

Then run:

Local setup

Development

Build:

Run the CLI from the project workspace:

Run tests:

CLI

crawl

docrawl crawl <url> [options]

Examples:

# Crawl a docs section into ./output
docrawl crawl https://docs.example.com/guide/

# Run a smaller smoke test first
docrawl crawl https://docs.example.com/guide/ --max-pages 10 --depth 1 --verbose

# Merge everything into one file
docrawl crawl https://docs.example.com/guide/ --single-file --output ./context.md

# Crawl the full hostname, not only the seed path subtree
docrawl crawl https://docs.example.com --domain --max-pages 200

Options:

-o, --output <path>  Output directory or file path
-s, --single-file    Merge all pages into one Markdown file
    --domain         Crawl the whole hostname, not just the seed path
    --depth <n>      Maximum crawl depth
    --max-pages <n>  Maximum pages to process (default: 500)
    --concurrency <n> Concurrent requests (default: 3)
    --delay <ms>     Delay between requests per worker (default: 500)
    --lang <code>    Preferred language for extraction, BCP 47
    --no-sitemap     Disable sitemap discovery
    --include <glob> Include URL glob pattern, repeatable
    --exclude <glob> Exclude URL glob pattern, repeatable
    --verbose        Detailed progress logging

By default, crawl prints one compact progress line per completed page. --verbose enables more detailed queue, fetch, discovery, and completion logs.

parse

docrawl parse <url> [options]

Examples:

# Parse one page as Markdown
docrawl parse https://docs.example.com/guide/intro

# Parse one page as JSON
docrawl parse https://docs.example.com/guide/intro --json

Options:

-j, --json      Output full JSON response
    --lang <code> Preferred language for extraction, BCP 47

--lang is passed to the extraction layer as a preference. It does not currently restrict crawling to URLs or pages of a specific language.

Output

Separate files

By default, docrawl crawl writes one Markdown file per successful page and a manifest.json.

Example layout:

output/
├── getting-started/
│   ├── introduction.md
│   └── quickstart.md
└── manifest.json

Each Markdown file includes frontmatter with fields such as:

• title
• sourceUrl
• finalUrl
• canonicalUrl
• crawledAt
• depth
• wordCount
• contentHash

Single file

With --single-file, docrawl writes:

• one merged Markdown file
• one adjacent manifest file named like <name>.manifest.json

The merged file includes a table of contents and one section per successful page.

Example:

docrawl crawl https://docs.example.com --single-file --output ./context.md

Produces:

context.md
context.manifest.json

Current limitations

docrawl currently does not handle:

• JavaScript-rendered SPAs that need browser execution
• login-gated or authenticated content
• asset downloading
• robots.txt compliance
• resumable crawls
• incremental recrawls
• full navigation reconstruction