For every parse, we return clean Markdown plus Litmus: 20 deterministic checks on who wrote the page, how fresh it is, and whether it’s even the page you asked for. Your agent doesn’t just read the web, it knows what to trust.
NanoParse renders the page (JavaScript included), strips every piece of boilerplate, and returns clean Markdown — plus the Litmus verdict your agent uses to decide whether to cite it.
<html> <head> <script>!function(f,b,e,v,n,t,s){if(f.fbq)return;n=f.fbq=function() {n.callMethod?n.callMethod.apply(n,arguments):n.queue.push(arguments)};if( !f._fbq)f._fbq=n;n.push=n;n.loaded=!0;n.version='2.0';n.queue=[];t=b.creat eElement(e);t.async=!0;t.src=v;s=b.getElementsByTagName(e)[0];s.parentNode. insertBefore(t,s)}(window,document,'script','https://connect.facebook.net/en _US/fbevents.js');</script> <link rel="stylesheet" href="/bundle.7f8a3c.css"> </head> <body class="template-article"> <div class="top-banner-ad"> <iframe src="https://ad.doubleclick.net/..."></iframe> </div> <nav id="sticky-header"> <ul><li><a href="/">Home</a></li> <li><a href="/sports">Sports</a></li> <li><a href="/opinion">Opinion</a></li></ul> </nav> <aside class="sidebar"> <div class="trending-widget">Most Read</div> <div class="newsletter-cta">Sign up for our newsletter</div> </aside> <div id="content"> <div class="article-header"> <h1>Study Finds Correlation Between Diet and Longevity</h1> </div> <p>A comprehensive 20-year study published in <em>The Lancet</em> has found a strong correlation between dietary patterns and increased lifespan.</p> <table class="data-table"> <thead><tr><th>Diet</th><th>Avg Lifespan</th><th>Sample</th></tr></thead> <tbody><tr><td>Mediterranean</td><td>83.4 yr</td><td>n=12,400</td></tr> <tr><td>Standard Western</td><td>76.1 yr</td><td>n=11,800</td></tr></tbody> </table> <div class="in-article-ad"> <img src="/ad-300x250.jpg" alt="Sponsored"> </div> </div> <footer> <div class="cookie-banner">We use cookies to improve your experience.</div> <div class="related-articles">You may also like...</div> <div class="comments-section">42 Comments</div> </footer> </body> </html>
# Study Finds Correlation Between Diet and Longevity *By Jane Chen, Health Correspondent* A comprehensive 20-year study published in *The Lancet* has found a strong correlation between dietary patterns and increased lifespan. | Diet | Avg Lifespan | Sample | | ----------------- | ------------ | --------- | | Mediterranean | 83.4 yr | n=12,400 | | Standard Western | 76.1 yr | n=11,800 | Researchers noted that participants following a Mediterranean diet showed a 7.3-year increase in average life expectancy compared to the control group. The study controlled for socioeconomic factors and exercise habits across 24,200 participants in seven countries.
{
"source": { "type": "news", "authority": 0.82 },
"freshness": { "published": "2026-01-14", "age": "7 months ago" },
"tldr": "20-year study links Mediterranean diet to a 7.3-year increase in average lifespan.",
"structural_trust_score": { "level": "high", "trust_factors": ["High-authority source"] },
"content_type": "news_article"
}
No account, no wallet — three free tries per IP per day, then $0.005 per parse via x402.
Extraction solved the text problem and left the judgement problem alone. Each of these is something an agent gets wrong today — confidently, and in writing.
No publish date, no staleness flag. A 2019 explainer reads exactly like last week's, and the citation looks identical either way.
A login wall or a truncated teaser passes as the article. The agent summarises the first eighty words of a page it never actually read.
A soft 404, a captcha, a "please enable JavaScript" shell — all arrive with a 200 status and get treated as a source.
Thin pages, syndicated reposts and boilerplate-heavy layouts rank like primary sources, because nothing measured the difference.
Every one is deterministic — no model in the path, so the same page returns the same signals every time and you can cite them in your own logs.
Who wrote it, what it claims to be, who it links to, and whether it shows its work.
05 SIGNALSHow much of the page is substance rather than hedging, furniture or assertion.
06 SIGNALSIs it current, is it reachable, is it gated, is it the original, is it the page you asked for.
03 SIGNALSWhat it says and how long it takes — enough to triage before you spend context.
No account, no API key, no subscription. Your agent pays per parse, machine to machine, and is charged only when the page renders.
HTTP 402, reserved since 1992, finally doing its job. Your agent gets payment terms, signs a USDC authorization, and retries — no checkout page, no dashboard, no human in the loop.
Payments settle in USDC on Base from a self-custodied wallet your agent holds. No seed phrases, no funds passing through NanoParse — the facilitator covers the network cost.
One flat price, one response. Markdown and Litmus come together — no second endpoint, no second charge — and you're billed only when the page renders. No gas fees. You pay one number: $0.005.
Straight answers for developers wiring up NanoParse.
Yes. Standard scrapers fail on modern SPAs because they only fetch raw HTML. NanoParse spins up a headless Chromium instance at the edge, fully executes all JavaScript, and waits for dynamic content to load before running content isolation algorithms.
For x402 payments, yes — the agent signs each payment from a self-custodied wallet holding
USDC on Base. If your agent doesn't have one, run npx @coinbase/payments-mcp:
the agent gets a Coinbase wallet, funds it via Apple Pay, and pays for services automatically.
Self-custody via Coinbase — no seed phrases, and nothing runs through NanoParse. You also get
10 free parses per month by signing with your wallet (no payment), or 3 anonymous free tries
per network.
An open payment protocol backed by Coinbase and Cloudflare. Uses HTTP 402 — reserved since 1992 — to let machines pay each other. Agent gets a 402 with payment terms, signs a USDC transaction, retries.
Structured intelligence about every page — source authority, freshness, structural trust score, hedge-language density, content density, content type, syndication status, key figures, reading time, and a TL;DR. Computed at the edge, no LLM call, zero extra cost. Learn more →
Firecrawl sells to humans: accounts, plans ($16–$599/mo), API keys. NanoParse is agent-native: no account, no plan. Agent discovers endpoint, pays per-call, moves on. See full comparison →
Yes. MCP server at nanoparse.app/mcp. Add to config, agent auto-discovers the tool.
Quick Start →
There are no gas fees. Your total cost is $0.005 per parse. Your wallet signs an
off-chain EIP-3009 authorization (the x402 protocol) and the payment facilitator
covers the network cost. You pay one number: $0.005.
You're never charged for a parse that doesn't succeed. Blocked, invalid, or abusive URLs are rejected before any charge — those never cost anything. And settlement only happens after a page renders successfully, so a failed render is simply never charged at all. Recently parsed URLs return instantly from cache, and retrying with the same signed payment is free.