llm.codes

llm.codes is a web service for converting JavaScript-heavy documentation into clean Markdown, optimized for language models.

Website: llm.codes Repository: github.com/amantus-ai/llm-codes

Modern documentation often uses:

• Client-side rendering (React, Vue)
• Dynamic content loading
• JavaScript navigation

AI assistants (Claude Code, Cursor) cannot execute JavaScript and see empty pages.

llm.codes:

1. Loads pages via headless browser (Firecrawl)
2. Waits for JavaScript rendering
3. Extracts main content
4. Converts to clean Markdown
5. Removes navigation, ads, boilerplate

Total: 69+ documentation sites

1. Open llm.codes
2. Paste documentation URL
3. Configure crawl depth
4. Click “Convert”
5. Copy Markdown

curl "https://llm.codes/api/scrape" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://docs.python.org/3/library/asyncio.html",
    "depth": 1
  }'

• 0 — only the specified page
• 1 — page + direct links
• 2 — recommended for most cases
• 5 — full documentation section

llm.codes removes:

• Navigation and sidebars
• Headers and footers
• Ads and banners
• Duplicate content
• Compatibility strings (iOS 14.0+, etc.)

Size reduction up to 70% while preserving all useful information.

• Results cached for 30 days
• Redis-backed for speed
• Repeat requests are instant

# Get documentation
curl -s "https://llm.codes/api/scrape?url=https://docs.example.com" \
  | claude "Explain how to use this API"

Below is the documentation for the XYZ library:

[paste llm.codes output]

Based on this documentation, help me implement...

git clone https://github.com/amantus-ai/llm-codes
cd llm-codes
npm install
npm run dev

FIRECRAWL_API_KEY=fc-...  # Required
UPSTASH_REDIS_URL=...     # Optional (for cache)

Apple uses heavy JavaScript. llm.codes is the only way to get content for LLM:

https://llm.codes/?url=https://developer.apple.com/documentation/swiftui

New React documentation (react.dev) also requires JS:

https://llm.codes/?url=https://react.dev/reference/react/hooks

• Only whitelist of 69 sites
• Requires Firecrawl API key for self-hosting
• Rate limiting for public API
• No support for authenticated content

llm.codes complements the llms.txt standard:

1. llms.txt — index of pages with links
2. llm.codes — get content of specific pages

Workflow:

llms.txt → select needed URLs → llm.codes → get Markdown