Link Preview & Open Graph API

Pay-per-result link preview API: $0.80 per 1,000 previews. Failed URLs are never charged.

Give it a list of URLs. Get back one row per URL with the title, description, og:image, all main Open Graph and Twitter card tags, the favicon, the canonical URL and the page language. Use it to build link cards, unfurl links in chat or a CMS, check social share tags, or enrich a URL list for an AI pipeline.

Also searched as: URL metadata API, Open Graph scraper, og:image extractor, link unfurl, URL preview, meta tags extractor, Open Graph checker.

Published by Rowfield Data. This is our own tool, built and operated with an AI agent.

Quick start

1. Run it with the prefilled input: {"urls": ["https://www.python.org/", "https://github.com/"]}. Two previews cost about $0.0017 (2 x $0.0008 plus the $0.00005 start).
2. Paste your own URLs (up to 5,000 per run, default cap 100 with maxUrls).
3. Export as CSV, JSON or Excel, or read the dataset through the API.

What you get per URL

• title, description (meta description)
• ogTitle, ogDescription, ogImage (absolute URL), ogImageAlt, ogType, ogSiteName, ogUrl, ogLocale
• twitterCard, twitterTitle, twitterDescription, twitterImage, twitterSite
• favicon (SVG or largest declared icon, else /favicon.ico), canonical, lang, themeColor
• finalUrl, statusCode, redirectCount, contentType, fetchedAt
• error and message: null for a preview; the reason for a free status row

What does it cost?

Pay per event: $0.0008 per preview ($0.80 per 1,000) plus $0.00005 per run start. Platform usage is included.

A preview is charged only when the page loads as HTML and has a title, description, Open Graph or Twitter tag. These rows are free:

• invalid URL or a private/local address
• blocked by robots.txt, or robots.txt not readable
• timeout, network error, HTTP 4xx or 5xx
• not an HTML page (PDF, image, JSON...)
• more than 3 redirects
• an HTML page with no title or meta tags at all

Examples: 1,000 URLs that all load: about $0.80. 1,000 URLs where 200 fail: about $0.64.

How it compares (prices checked 2026-11-02)

Read from each vendor's public pricing page or Apify Store pricing on 2026-11-02. Prices can change; check before you rely on them.

Service	Price model	Per 1,000
This actor	Pay per preview, no subscription	$0.80 (failed URLs free)
OpenGraph.io Developer	$25/month for 50,000 credits; overage $1.00 per 1,000	$0.50 if you use all credits, $1.00 overage
Microlink Pro	$49/month for 46,000 requests	about $1.07
thoob/url-metadata-extractor (Apify)	Pay per URL read	$1.00
technicaldost/link-preview-api (Apify)	Pay per result + $0.001 start	$2.00
mighty_monk/open-graph-extractor (Apify)	Pay per item + $0.00005 start	$3.00

If you send more than about 31,000 URLs every month ($25 / $0.80 per 1,000), a monthly OpenGraph.io plan can cost less per URL. This actor has no monthly fee, so small and irregular jobs pay only for what loads.

Input

• urls: page URLs, e.g. https://www.python.org/. A bare domain like python.org becomes https://python.org/.
• startUrls: the same as a URL list (combined with urls).
• maxUrls: only the first this-many unique URLs are read (default 100, max 5,000).
• maxConcurrency: how many different hosts are read at once (default 5, max 10).

Output example

{
  "rowType": "preview",
  "url": "https://www.w3.org/",
  "finalUrl": "https://www.w3.org/",
  "statusCode": 200,
  "title": "W3C",
  "description": "The World Wide Web Consortium (W3C) develops standards and guidelines...",
  "ogTitle": "W3C",
  "ogImage": "https://www.w3.org/assets/website-2021/images/w3c-opengraph-image.png",
  "ogType": "website",
  "ogSiteName": "W3C",
  "twitterCard": "summary_large_image",
  "error": null
}

A status row has rowType status, the reason in error (for example robots_disallowed or http_4xx) and a plain note in message.

Polite access

• robots.txt is read first for every host, including each new host in a redirect chain. Disallowed pages are not fetched. A robots.txt that answers with an error other than 404/410 means the host is skipped (free row).
• One request at a time per host, at least 1 second apart (longer if the site sets a Crawl-delay, up to 10 seconds).
• Plain HTTP GET only: no browser, no cookies, no proxy, no login. The user agent names this actor.
• 15-second timeout, at most 3 redirects, reads at most 1 MB and stops after </head>.
• Private, loopback and local network addresses are refused.

Privacy

Only page-level metadata is returned. Author tags (author, article:author, twitter:creator), e-mail addresses, phone numbers and person names are not collected.

Use with AI agents (MCP)

Add this actor as a tool in an MCP client through Apify's MCP server:

https://mcp.apify.com?tools=founder-experiment/link-preview-open-graph-api

Your agent can then ask for "the title and og:image of these links" and get one row per URL. Set maxUrls in the tool call to cap cost.

Related tools

Other actors from the same publisher (Rowfield Data):

• Company Logo & Favicon: best logo URL, favicons and app icons for a list of domains.
• Sitemap URL Extractor: every URL from a website's sitemaps, ready to feed into this actor.

FAQ

Does it render JavaScript? No. It reads the HTML the server sends. Sites that add their meta tags with JavaScript only may return fewer fields.

Why did a big site fail? Some sites block automated requests or ask bots not to crawl in robots.txt. You get a free status row, and you are not charged.

Is the og:image checked? The URL is returned as the page declares it (made absolute). It is not downloaded.

Limits

• Up to 5,000 URLs per run. Split bigger lists into several runs.
• Values are kept as the page publishes them (trimmed and shortened to safe lengths).
• Built and operated by an AI agent. Report problems in the Issues tab.

Changelog

• 2026-11-02: First version, pay per preview ($0.0008 per preview, $0.00005 per run start).