---
name: web-extract
description: |
  Turn any public URL into clean, LLM-ready data in one paid call (x402, USDC on Base) — structured JSON
  or Markdown. No API keys, no signup, pay-per-call. Handles redirects, timeouts, size caps, and SSRF safely.

  USE FOR:
  - Extracting a page to structured JSON: title, description, full text, ALL JSON-LD, OpenGraph/Twitter meta, headings, links
  - Reading a page as clean Markdown for LLM context (nav/ads/scripts stripped; headings/links/lists preserved)
  - Reliable web-reading before feeding a page to a model (safer than raw fetch: SSRF guards, size limits)

  TRIGGERS:
  - "extract data from <url>", "get the structured data / JSON-LD from <url>", "scrape <url>"
  - "read <url> as markdown", "give me the content of <url> for context", "summarize this page" (fetch first)
metadata:
  version: 1
---

# Web Extract / Read (URL -> structured data or Markdown) via x402

Two endpoints, one skill. Public data only, **no auth, no API keys, no subscription — pay per request in USDC on Base.**

| Task | Endpoint | Price | Param |
|------|----------|-------|-------|
| URL -> structured JSON (text, JSON-LD, OG, headings, links, AI-readiness) | `GET /extract?url=` | **$0.05** | `url` |
| URL -> clean LLM-ready Markdown | `GET /read?url=` | **$0.05** | `url` |

Base host: `https://x402-url-extractor-production.up.railway.app` (alias `https://pay.samedaydesk.com`). Network: USDC on Base mainnet (`eip155:8453`).

## How to pay (x402)

`GET` the URL → **HTTP 402** with payment requirements → pay with any x402 client (`x402-fetch`, `x402-axios`,
Coinbase **AgentKit** x402 action) from a USDC-funded Base wallet → client replays with `X-PAYMENT` → JSON result.

```ts
import { wrapFetchWithPayment } from "x402-fetch";
const fetchPaid = wrapFetchWithPayment(fetch, walletClient);
const res = await fetchPaid("https://x402-url-extractor-production.up.railway.app/extract?url=https://example.com");
const data = await res.json();
```

## Output (extract, abridged)

```json
{ "ok": true, "url": "https://example.com", "title": "...", "description": "...",
  "jsonLd": [], "openGraph": {}, "headings": {}, "links": [], "text": "...",
  "aiReadiness": { "hasJsonLd": false, "schemaTypes": [] } }
```

`/read` returns `{ ok, url, title, markdown, wordCount, truncated }`.

## Notes

- For domain/company intelligence specifically, prefer the `company-enrich` skill (richer firmographics + DNS + AI-readiness).
- Both handle redirects, timeouts, size caps and SSRF safely — use them instead of raw `fetch` for untrusted URLs.