Text Extraction API
Extract plain text from any URL for NLP, search indexing, and LLM pipelines. No HTML tags, no scripts, no parsing to write.
HTML isn't text
HTML is full of tags, scripts, and markup your NLP or search pipeline doesn't need
Stripping it yourself means writing and maintaining a parser
Nav bars, ads, and footers still leak into your word counts and embeddings
Clean plain text
Lowest token formatformat: "text"returns the full page as raw text, ~70% fewer tokens than raw HTMLformat: "text-llm"additionally strips nav and boilerplate, ~85% fewer tokens than HTMLNo parser to write or maintain, one request, plain text back
Simply use the format: "text" (or "text-llm") parameter on our Web Scraping API.
Built for text-only pipelines
Boilerplate stripped
Use format: "text-llm" to drop navigation, ads, footers, and sidebars, leaving only the primary content.
Lowest token cost
Text carries no markup overhead, so it's cheaper on tokens than Markdown or HTML for the same page.
LLM and NLP ready
Feed straight into sentiment analysis, embeddings, full-text search indexing, or any pipeline that just needs words.
What developers build with plain text
Text is the leanest format for NLP, search, and LLM context.
Scrape static documentation and knowledge bases without browser overhead. Output clean text to minimize token usage for LLMs.
{
"url": "https://www.twilio.com/docs/usage",
"format": "text-llm"
}Scrape static documentation and knowledge bases without browser overhead. Output clean text to minimize token usage for LLMs.
{
"url": "https://www.twilio.com/docs/usage",
"format": "text-llm"
}Frequently Asked Questions
text returns the rendered page as raw text, tags stripped but boilerplate still included. text-llm goes further and strips that boilerplate too, leaving only the primary content.
Yes. Text carries no markup syntax at all, so for the same page it's smaller than the equivalent Markdown output. If you don't need headings, links, or table structure preserved, text is the leanest option.
No. Both text and text-llm are fully flattened, plain words with no markup, no distinct headings or list structure. If you need structure preserved, use format: "markdown" instead.
Yes, that's a common use case. Feed the plain text output straight into Elasticsearch, Algolia, or a vector database.
Text Extraction returns only the flattened text content. HTML Extraction returns the full rendered DOM for you to parse with your own selectors. Use text when you just need the words, HTML when you need to target specific elements.