Agent Read Proxy

A web scraping API for AI agents: give it a web page address, get back clean text or Markdown plus page metadata. No API key or signup — agents pay per call in USDC via x402.

Price

$0.002 / page

Payment

x402 · USDC

Network

Base

How to call it

GET https://agent-read-proxy.agent-read-proxy.workers.dev/v1/read?url=<https URL>&format=markdown

format is optional: text (default) returns plain text, markdown keeps headings, lists and links. Every response also includes the page's title and, when the page provides them, its description, author, publish date, site name, language, canonical URL and preview image.

The first request answers 402 Payment Required with x402 payment details. An x402 client pays and retries automatically. If the page can't be fetched or isn't a web page, you get an error and aren't charged.

Example

GET https://agent-read-proxy.agent-read-proxy.workers.dev/v1/read?url=https://en.wikipedia.org/wiki/Motocross&format=markdown

{
  "url": "https://en.wikipedia.org/wiki/Motocross",
  "title": "Motocross - Wikipedia",
  "language": "en",
  "canonical_url": "https://en.wikipedia.org/wiki/Motocross",
  "format": "markdown",
  "markdown": "From Wikipedia, the free encyclopedia\n\nOff-road motorcycle racing held on enclosed off-road circuits\n\n## History...",
  "word_count": 4556,
  "truncated": false,
  "fetched_at": "2026-10-28T01:30:00.000Z"
}

Use cases

Research agents that need to read articles, docs and blog posts from search results
RAG pipelines that ingest web pages as clean Markdown for embeddings and retrieval
News and change monitoring: pull the latest text of a page and compare it over time
Price and product research: extract readable product and pricing pages
Summarization and Q&A: give an LLM the page's real content without ads, menus or scripts
Citations: get the canonical URL, title, author and publish date along with the text

What it strips out

Scripts, styles, menus, headers, footers, sidebars and hidden elements
When a page marks its main content, only that is returned

Limits

Public http/https pages on standard ports only; private and local addresses are refused
Up to 2 MB downloaded, 10 second timeout, 50,000 characters of text
HTML and plain-text pages (no images, PDFs or downloads)