Let your AI agent read any webpage, cheaply.
The Canonical x402 Reader™
Skim™ turns any URL into agent-ready markdown — or typed JSON if you pass a schema. No account, no API keys — your agent calls, pays in USDC over x402, and gets clean content back. $0.002 per read, $0.015 per structured extract.
Don't pay LLM rates for non-LLM work.
Smart agents delegate — same work, a fraction of the energy.
10 free skims a day for humans — no signup, no wallet, no API key.
Consistently faster than Firecrawl on a 15-URL production benchmark. Every call on-chain.
Clean markdown comes back ~4x smaller than the raw HTML — with nothing important lost.
Faster here too — see for yourself.
Skim calls live from your browser.
Firecrawl's number is from our recorded benchmark on the performance page.
Skim's number is measured live in your browser against /api/demo/race. Firecrawl's number is the recorded result from our 2026-05-22 benchmark on the same URL — we don't hit their API per visitor (API key + cost). Rate-limited to 20 runs/min per IP.
See the full 15-URL benchmark — every call backed by an on-chain transaction on Base.
Cheap because it's junk? No — cheap because it's x402.
$0.002 per read isn't a bargain-bin price for a bargain-bin service. It's what clean web reading costs in the new x402 agentic economy — no accounts, no sales teams, no LLM overhead, settled per call in USDC. Better performance for a lot less.
Example: 10,000 pages a month, estimated at public token list prices. Volume discount starts automatically at 50,000 reads per month.
Calculate your savingsYes — you can vibe code this.
New to all this? You don't need to understand a single line of the code on this page. If you're building with an AI agent, just tell it: “set me up to use Skim so you can read web pages for me.” Point it at this website and it handles the rest — creating a wallet, paying per call, and wiring everything up while you watch. The code samples below are for engineers; you can skip them entirely.
Point your agent at skim402.com/docs and ask it to set up clean web reads with Skim.
One HTTP call.
Your agent POSTs a URL. Skim returns 402 with payment instructions, the agent pays a USDC micropayment, and on retry gets clean markdown + structured metadata back.
curl -X POST https://skim402.com/api/v1/read \
-H "Content-Type: application/json" \
-d '{ "url": "https://example.com/article" }'
# 402 Payment Required → pay with x402 client → retryAgent-native payments
Built on x402 — your agent pays per request in USDC over HTTP. No accounts to provision, no keys to rotate, no card on file.
Markdown or typed JSON
Nav, ads, scripts, cookie banners — gone. Get pure article markdown your LLM can use (~4x smaller than the raw HTML, with nothing important lost), or pass a JSON Schema to /v1/extract and get back typed JSON guaranteed to match. Six built-in presets — article, product, job, review, event, table — need no schema at all.
Fast & safe
Sub-second responses for static pages, with SSRF protection built in so an autonomous agent can't be tricked into hitting internal infrastructure.
The rail Coinbase is building its agent stack on.
In June 2026 Coinbase launched Coinbase for Agents — letting an AI agent trade and pay straight from a Coinbase account — and said it will "soon be x402-enabled." x402 is the exact standard Skim runs on today. Your agent doesn't have to wait for agentic payments to arrive; it can pay for a clean web read right now.
Why this matters for x402
Why agents pick Skim over a traditional reader API
Traditional reader APIs
- ✗ Human signs up, adds a card, copies a key
- ✗ Monthly plans force prediction of usage
- ✗ Agents can't authenticate themselves
- ✗ Quota limits block at the worst time
Skim
- Agent calls a URL, pays, gets content
- $0.002 per read, $0.015 per structured extract
- Failed reads cost nothing — pay only on 2xx
- Native HTTP 402 — works with any x402 client
- No quotas — pay