Skip to main content
Reader can return the same scrape as multiple formats in one request. You pay for one scrape (1 or 3 credits depending on mode), and the response carries whichever formats you asked for.

​
Request both

const result = await client.read({
  url: "https://example.com/blog/post",
  formats: ["markdown", "html"],
});

if (result.kind === "scrape") {
  console.log(result.data.markdown); // for the LLM
  console.log(result.data.html);     // for custom DOM parsing
}

​
When to use markdown only

The default. Clean, tokenizer-friendly, good for LLMs and RAG. Use markdown by itself unless you know you need HTML.

​
When to include HTML

  • You need structure Reader’s markdown conversion strips. For example, <table>s with complex layouts sometimes lose nuance in markdown. The HTML preserves it.
  • You want to run your own parser. If you already have a cheerio / BeautifulSoup pipeline and just need Reader’s clean HTML as input, skip the markdown.
  • You’re extracting specific elements. You want just the first <img>, or every <blockquote>, without parsing markdown back.
Asking for both formats costs the same as asking for one: Reader scrapes once and serializes twice. Leave both on during development, drop to ["markdown"] in production once you know what you need.

​
What HTML you actually get

The HTML Reader returns is cleaned, not the raw DOM. Scripts, styles, tracking pixels, and (if onlyMainContent: true) boilerplate are already removed. It’s the same DOM Reader used internally to generate the markdown: good for parsing, not for rehydrating the original page. If you need the absolute raw HTML, set onlyMainContent: false. Reader will skip its boilerplate stripping and give you closer to the source.

​
Next