Extract Clean Article Text from Any URL

Most web pages are 80% boilerplate: navigation, footers, ads, cookie banners, and sidebars. When you need the article itself, a screenshot gives you pixels and a raw fetch gives you a wall of markup. The extract endpoint returns just the readable content: title, author, excerpt, and body in Markdown, JSON, or HTML, drawn from a full Chromium render pass. SPAs and JavaScript-heavy pages work the same as static ones.

This guide shows how to use the extract article text API from curl, the Python SDK, and the JavaScript SDK, and how to combine extraction with screenshot capture in a single request.

Prerequisites

A Rendex API key. The free tier gives you 100 calls per month with no credit card required.
curl (for the HTTP examples), or the Python or JavaScript SDK.

Step 1: Send your first extraction request

POST /v1/extract takes a URL and returns the article body without images, navigation, or ads.

extract.sh

curl https://api.rendex.dev/v1/extract \
  -H "Authorization: Bearer rdx_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://example.com/blog/article",
    "extractFormat": "markdown"
  }'
# Get your API key at rendex.dev/login

The response includes the extracted content and page metadata:

response.json

{
  "success": true,
  "data": {
    "url": "https://example.com/blog/article",
    "format": "markdown",
    "content": "# Article Heading\n\nFirst paragraph of clean body text...",
    "title": "Article Heading",
    "byline": "Author Name",
    "excerpt": "A short description of the article content.",
    "siteName": "Example",
    "length": 4218,
    "loadTimeMs": 580
  }
}

length is the character count of the extracted plain text. Use it to estimate token cost before feeding the content to a language model. loadTimeMs tracks render time, which varies by page weight.

Step 2: Choose the right extractFormat

The extractFormat parameter controls the shape of the content field. Default is markdown.

markdown: Headers, bold, links, and code blocks are preserved as Markdown. Paste directly into an LLM prompt or a document store.
json: The full Readability parse result as a JSON string. Includes additional fields beyond what the API surfaces at the top level.
html: Sanitized HTML with semantic structure. Useful when you render the content in a frame or need to preserve anchor IDs.
text: Flat plain text with no markup. Smallest output, fastest to process, good for embedding pipelines that do not need structure.

Step 3: Extract with the Python SDK

extract.py

# pip install rendex
from rendex import Rendex

rendex = Rendex("rdx_YOUR_KEY")  # get your key at rendex.dev/login

result = rendex.extract(
    "https://example.com/blog/article",
    extract_format="markdown",
)

print(result["title"])    # "Article Heading"
print(result["byline"])   # "Author Name"
print(result["length"])   # 4218
print(result["content"])  # full Markdown body text

The extract method posts to /v1/extract. The extract_formatparameter converts to the API's camelCase extractFormat automatically.

Step 4: Extract with the JavaScript SDK

extract.ts

import { Rendex } from "@copperline/rendex"

const rendex = new Rendex("rdx_YOUR_KEY")  // get your key at rendex.dev/login

const result = await rendex.extract({
  url: "https://example.com/blog/article",
  extractFormat: "markdown",
})

console.log(result.title)    // "Article Heading"
console.log(result.byline)   // "Author Name"
console.log(result.length)   // 4218
console.log(result.content)  // full Markdown body

Step 5: Feed extracted text to an LLM

The extract endpoint strips all the noise before you call a language model. No HTML cleanup, no wasted tokens on menus.

summarize.py

import anthropic
from rendex import Rendex

rendex = Rendex("rdx_YOUR_KEY")
claude = anthropic.Anthropic()

result = rendex.extract(
    "https://example.com/blog/article",
    extract_format="text",  # plain text uses fewer tokens
)

message = claude.messages.create(
    model="claude-sonnet-5",
    max_tokens=512,
    messages=[{
        "role": "user",
        "content": f"Summarize in 2-3 sentences:\n\n{result['content']}",
    }],
)

print(message.content[0].text)

Step 6: Get a screenshot and text in one call

To capture an image and extract the article body together, use POST /v1/screenshot/json with extract: true. The response includes the base64-encoded image alongside an extracted key. Extraction on this endpoint is non-fatal: if the page has no article-like content, extracted is null and the screenshot proceeds normally.

screenshot-and-extract.sh

curl https://api.rendex.dev/v1/screenshot/json \
  -H "Authorization: Bearer rdx_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://example.com/blog/article",
    "extract": true,
    "extractFormat": "markdown"
  }'

Example: extraction report

Article extraction report showing extracted title, author, and clean Markdown content from a URL using the extract article text API — A Rendex Article Extraction Report showing the title, author, excerpt, and Markdown body returned by POST /v1/extract

Troubleshooting

EXTRACTION_FAILED (422). The page has no article-like content that the extractor can identify. Landing pages, dashboards, and login screens fall into this category. Use the free URL-to-Markdown tool to test a URL before building against it.

Async is not supported. /v1/extract returns content inline and rejects async: true. For high-volume extraction, manage your own job queue or use the batch screenshot endpoint with extract: true.

Geo-targeting is not available on /v1/extract. To extract content from a geo-restricted URL, use POST /v1/screenshot/json with both geo and extract: true.

HTML input is not supported. /v1/extract requires a url. To extract from your own HTML string, use POST /v1/screenshot/json with html and extract: true.

Next steps

Test extraction on any URL without writing code using the free URL-to-Markdown tool. The API reference covers every parameter the extract endpoint accepts, including cookie and header injection for pages that require authentication. For a broader look at where extraction fits in the rendering stack, see Rendering API vs Screenshot API.

Get a free API key and start pulling clean article text from any URL in seconds.