Unrender

Rendered page → clean Markdown.
Runs a real browser (JS included) and uses screenshot-grounded LLM extraction to return only the content a human would read.

No "Skip to content." No cookie banner sludge. No div soup. Just stable, link-preserving Markdown for RAG, notes, and pipelines.

Unlock your agent to read the JS-heavy human web

• LLM extraction — well structured outputs with understanding.
• Real browser rendering — works on SPAs/Next.js/React.
• Direct agent integration — drop a snippet into your CLAUDE.md and go.

Get an API token

Try it out

• Apple CallKit Docs
• Simon Willison Blog
• Karpathy on X

Or enter a URL:

{CLAUDE,AGENTS}.md

## Fetching Web Pages as Markdown

Use unrender if builtin webfetch didn't return good results,
because of JS-heavy implementation.

  # Fetch a webpage as Markdown
  curl -H "Authorization: Bearer YOUR_TOKEN" \
    https://unrender.page/https://example.com

Free tier: 5 req/min, 200 req/day.