URL to Markdown - Clean Web Page Text for LLMs
Pricing
$2.00 / 1,000 page converteds
URL to Markdown - Clean Web Page Text for LLMs
Turn any web page into clean Markdown and plain text for LLMs, RAG and summaries. Finds the real article, drops menus, cookie banners and scripts, keeps headings, lists, links, images, quotes and code, and returns title, author, date and language. No browser. Failed pages are free.
Pricing
$2.00 / 1,000 page converteds
Rating
0.0
(0)
Developer
Scrapewise Data
Maintained by CommunityActor stats
0
Bookmarked
2
Total users
1
Monthly active users
2 days ago
Last modified
Categories
Share
URL to Markdown: clean web page text for LLMs
Give it a list of page URLs and get back clean Markdown and plain text, ready to paste into a prompt, index in a RAG pipeline or store as a document. No browser, no API key, no HTML cleanup on your side.
What comes out
| Field | What it holds |
|---|---|
markdown | headings, lists, links, images, quotes and code blocks, with every link absolute |
text | the same content as plain text, for embeddings and search |
wordCount, charCount, headingCount, linkCount, imageCount | size and shape of what was found |
title, description, author, publishedAt, language, siteName, imageUrl, canonicalUrl | the metadata a document pipeline usually wants |
finalUrl, statusCode, htmlBytes, renderedByJavaScript | what actually happened when fetching |
Why it is not just "strip the tags"
- It finds the article.
<article>,<main>,[role=main], or the densest block of paragraphs, and drops navigation, footer, sidebar, cookie banner, script and style. - It survives minified HTML. Attributes without quotes (
href=/docs class=x) are common in production builds and break naive parsers; here they are handled. - It tells you when a page is JavaScript-only.
renderedByJavaScriptis true when the HTML is big but the text is tiny, so you know the page needs a browser instead of getting a stub and not noticing. - You can set a floor.
minWordsturns anything thinner into a free error row, so a broken page never enters your dataset as if it were content.
Errors, all free
| errorCode | Meaning |
|---|---|
NOT_FOUND | 404 or 410 |
BLOCKED | 401 or 403: the site refused. Turn the proxy on and try again. |
FETCH_FAILED | DNS, timeout or a server error that did not clear on retry |
TOO_LITTLE_TEXT | The page came back with less text than your minWords |
ITEM_UNREADABLE | The page came in a shape this Actor cannot read |
UNEXPECTED | Something unforeseen broke on that page; the rest of the run still delivers |
Pricing
US$ 2.00 per 1,000 pages, pay per event: no monthly fee, no charge per run, error rows free. A thousand pages for a RAG index cost US$ 2.00.
FAQ
How is this different from stripping the tags myself? The Actor picks the content block
(<article>, <main>, [role=main] or the densest block of paragraphs) and drops navigation,
footer, sidebar, cookie banner and script, so what you get is the article and not the whole page.
What if the page only renders with JavaScript? You still get whatever is in the HTML, and
renderedByJavaScript is set to true so you can route those URLs to a browser-based Actor
instead of silently indexing a stub.
Can I refuse thin pages? Yes. Set minWords and anything below it comes back as a free
TOO_LITTLE_TEXT row instead of entering your dataset as content.
Are the links usable? Yes, every link and image is made absolute against the final URL, after redirects.
Do failed pages cost anything? No. 404, 403, timeouts and thin pages are all free rows.
Notes
- Proxy is off by default and the platform cost is a fraction of the price. Turn it on for sites that block datacenter IPs.
- This Actor does not run JavaScript. For pages that only render in a browser,
markdownwill be short andrenderedByJavaScriptwill be true.
