URL Metadata & Link Preview API (no key)

Turn any URL into a clean link preview — behind a simple HTTP API (and MCP tool). Extract OpenGraph, Twitter Card and standard meta (title, description, image, site name, favicon, canonical, theme color, language, JSON-LD), or list a page's internal / external links and images. No browser, no key — just fast HTML parsing. Pay per query.

• 🔗 Link preview — title, description, preview image, favicon, site name, type, canonical URL, article:published_time, JSON-LD structured data.
• 🗂️ Links & images — every link on a page, resolved to absolute URLs, deduped and split into internal vs external, plus image URLs and counts.
• 🛡️ Safe by design — only public http(s) URLs; localhost, private, link-local and cloud-metadata addresses are refused.
• ⚡ No browser — plain fetch + Cheerio: fast and cheap.
• 💵 Pay per query.

The use case it was built for

Build rich link cards for a chat app or CMS, prefill an "unfurl" preview, audit a page's outbound links, or hand your agent a clean summary of what's at a URL — one call in, structured JSON out.

Sample output (metadata, url "https://github.com")

{
  "ok": true, "finalUrl": "https://github.com/", "status": 200,
  "title": "GitHub · Build and ship software…",
  "description": "Join the world's most widely adopted…",
  "siteName": "GitHub", "type": "object",
  "image": "https://github.githubassets.com/…/og-image.png",
  "favicon": "https://github.com/favicon.svg",
  "canonical": "https://github.com/", "lang": "en", "twitterCard": "summary_large_image"
}

How to use

• HTTP API (Standby): POST or GET /metadata, /links. GET / returns help. Example: GET /metadata?url=https://news.ycombinator.com.
• Normal run: pass { operation, url, ... }; results land in the dataset + key-value store.
• MCP tool: expose it to your agent via Apify's MCP server.

Input

• operation — metadata · links.
• url — the page to inspect.
• limit — (links) cap the number of links. timeoutMs — fetch timeout.

Pricing

Per query from $0.004 (down to $0.001 on higher tiers). One call = one extraction.

Related Actors

• Website Screenshot & PDF — render a page to an image or PDF.
• Domain & DNS Lookup — DNS records, whois and IP info.

FAQ

Do I need a key? No.

Does it run JavaScript? No — it parses the server-rendered HTML (fast and cheap). For pages that need a real browser, pair it with our Website Screenshot actor.

Which private URLs are blocked? Anything non-public: localhost, 10.x, 192.168.x, 172.16–31.x, 169.254.x (incl. cloud metadata) and IPv6 equivalents.

Notes & limits

Parses up to ~5 MB of HTML per page. Some sites block automated requests or serve different HTML to bots. JSON-LD is returned as-is (up to 5 blocks).