Turn any URL into clean Markdown for your LLM
Raw HTML wastes an LLM's context on menus, scripts and styling. One call turns a web page, PDF, Word or Excel file into clean Markdown with its title and links, for $0.005.
Add it to Claude or Cursor free (30 seconds) See a live answer first
Why Markdown
Markdown keeps what a model needs (headings, lists, tables and links) and drops what it doesn't. A typical page shrinks to a fraction of its HTML size, so the model reads more of what matters and you pay for fewer tokens.
Which tool to use
- URL to Markdown ($0.005): web pages and documents: PDF, Word (.docx), Excel, OpenDocument, CSV, XML and text, up to 10 MB.
- Page extract ($0.005): a web page as Markdown plus every link as JSON.
- Site crawl ($0.03): up to 10 pages of a docs site or blog in one call, following its sitemap and obeying robots.txt.
- Structured extraction ($0.03): give your own JSON Schema and get the page's data back in that shape (prices, contacts, specs).
- Summarise ($0.02): a short or long summary of any page.
From code
Call it without paying to see the price, then pay per call with any x402 client or send an API key with prepaid credits; see x402 examples. Private and internal addresses are blocked, and failed fetches are never charged.
curl -i "https://aayatai.com/markdown?url=https%3A%2F%2Fexample.com"In Claude or Cursor
Add the MCP server and ask "Read this page and summarise the pricing section: https://…". Your assistant fetches the page as Markdown first.
Live example
Live example: a web page as Markdown
Free, no sign-up: a real answer from our own health check for this input. Your own questions cost $0.005 a call, and the first 20 a day are free in Claude or Cursor.
https://aayatai.com/markdown?url=https%3A%2F%2Fexample.com Ask your AI assistant
Once it's added, just ask in plain words. For example:
Read https://en.wikipedia.org/wiki/HTTP_402 as markdown and tell me what 402 is used for.Get a free key and add it All tools and prices
Tools on this page
4 of these 6 are in the free trial (20 calls a day over MCP). Prices come straight from our live price list; failed calls are never charged.
| Tool | What it answers | Price per call |
|---|---|---|
| Markdown | Turn any document URL into clean Markdown for LLMs: web pages, PDFs, Word (.docx), Excel (.xlsx/.xls), OpenDocument, CSV, XML and plain text, up to 10 MB. | $0.005free to try |
| Extract | Fetch any public web page and get back clean Markdown, the page title and all links as JSON. | $0.005free to try |
| Crawl | Turn a small website, or one section of it, into clean Markdown: starts from your URL, finds pages from the sitemap or by following links on the same site, obeys robots.txt, and returns up to 10 pages (title, Markdown, address) in one call. | $0.03free to try |
| Extract JSON | AI structured extraction: give a URL (web page, PDF, Word, Excel...) or text plus your own JSON Schema, get back JSON that fits it. | $0.03 |
| Summarise | Summarise text or any public web page with AI. | $0.02 |
| Metadata | A web page's metadata in one cheap call: title, description, canonical URL, language, site name, preview image, OpenGraph and Twitter card tags, author and dates, icons, RSS/Atom feeds and JSON-LD (schema.org) types. | $0.002free to try |
Questions people ask
Does it run JavaScript?
No: it reads the page as served, which works for most sites, articles and docs. Pages that only render in a browser may come back thin.
Does it respect robots.txt?
The site crawl obeys robots.txt. Single-page fetches read one public page you point to, like a browser would. Private and internal addresses are always blocked.
What does it cost?
$0.005 per page. In Claude or Cursor the first 20 calls a day are free, with no sign-up. Failed fetches are never charged.
Is there a size limit?
Documents up to 10 MB. Very slow or huge pages are rejected without charge.
Related
More: every use case · developer docs · llms.txt