URL to Clean Markdown — LLM-Ready Text Extractor

Turn any web page into clean, readable Markdown. Give it a URL, get back just the content — no ads, no navigation bars, no cookie banners, no scripts.

Built for feeding web content to AI models, where junk in the input means junk in the output.

---

Why use this

Most scrapers hand you raw HTML and leave the cleanup to you. This one does the cleanup:

• Readable output, not HTML soup. Uses Mozilla's Readability engine — the same technology behind Firefox's Reader View — to find the actual article and throw away the rest.
• Proper Markdown. Headings, lists, tables, code blocks, and links are all preserved in GitHub-flavoured Markdown, ready to paste into a prompt.
• Handles modern sites. Flip on Render JavaScript and the page loads in a real browser first, so content built by JavaScript is captured too.
• Batch friendly. Pass a list of URLs and get one clean record per page.
• You only pay for what works. Failed pages are recorded so you can see what happened — and you are never charged for them.

Great for

• Feeding articles and documentation into LLMs, RAG pipelines, and AI agents
• Building research and knowledge bases from web sources
• Content archiving in a durable, human-readable format
• Turning docs sites into clean Markdown for training or reference data

---

Input

Field	Type	Description
startUrls	array	One or more page URLs to convert. Each successful page is charged once.
url	string	Shortcut for a single page. Can be used instead of, or alongside, startUrls.
renderJs	boolean	Load each page in a real browser so JavaScript-built content is captured. Slower. Default false.
renderWaitMs	integer	With renderJs on, wait this many extra milliseconds after load for late content. Default 0.
includeLinks	boolean	Keep hyperlinks in the Markdown. Turn off for cleaner plain prose. Default true.

Example input

{
  "startUrls": [
    "https://en.wikipedia.org/wiki/Markdown",
    "https://example.com"
  ],
  "renderJs": false,
  "includeLinks": true
}

Output

One dataset record per page:

{
  "url": "https://en.wikipedia.org/wiki/Markdown",
  "title": "Markdown",
  "excerpt": "Markdown is a lightweight markup language...",
  "siteName": "Wikimedia Foundation, Inc.",
  "wordCount": 3092,
  "markdown": "Markdown is a lightweight markup language for creating formatted text...",
  "scrapedAt": "2026-07-23T20:31:00.000Z"
}

If a page can't be scraped, you get a record with an error field explaining why — and that page is not charged.

---

Pricing

Pay per event — you are charged once per page successfully scraped. No subscription, and failed pages are never charged.

What you run	Charged as
Standard scrape (default)	Page scraped — the low rate
With Render JavaScript on	Page rendered — a higher rate

Rendering runs a real browser and uses far more computing power, so it's priced separately. If you don't turn it on, you never pay the rendering rate.

---

When to turn on "Render JavaScript"

Leave it off by default — it's faster and works for most articles, blogs, documentation, and news sites.

Turn it on when a page comes back nearly empty. That usually means the site builds its content with JavaScript after loading. The difference on such a page is dramatic:

Mode	Result on a JavaScript-built page
Default (off)	~3 words — just navigation links
Render JavaScript (on)	~190 words — the full content

---

Good to know

• Only http and https pages are supported.
• Pages are capped at 5 MB and time out after 20 seconds (30 with rendering) to keep runs fast and predictable.
• Internal and private network addresses are refused for security.
• A small number of sites block automated traffic; that's a site policy, and those pages will be reported as errors rather than charged.