WebReader
Any web page in. Clean, AI-readable Markdown out.
WebReader is a pay-per-call service for AI agents. Agents can't read most web pages — they're buried in ads, popups, and layout markup. WebReader fetches the page, extracts the main content, and returns tidy Markdown an agent can actually use.
Endpoint
POST /api/read
Content-Type: application/json
{ "url": "https://example.com/some-article" }Response
{
"url": "https://example.com/some-article",
"title": "The Article Title",
"excerpt": "A short summary of the page…",
"markdown": "# The Article Title\n\nThe full article text…",
"wordCount": 842
}Research endpoint
Don't fetch pages one by one — ask the question and get the finished answer. The research endpoint searches the web, reads the top sources, and returns a briefing in Markdown with citations.
POST /api/research
Content-Type: application/json
{ "question": "What are the current US federal EV tax credit rules?" }The response carries the briefing (Markdown, with inline citations and a source list), the sources used, and a wordCount. Failed research runs are never charged.
Pricing
$0.10 per read and $1.00 per research call, paid in USDC on Base via the x402 protocol. No API key, no account, no subscription. Call without payment and you'll get an HTTP 402 with payment instructions; pay, retry, and the page is yours. Failed reads are never charged.