Metascraper Metadata Extractor

What is Metascraper Metadata Extractor?

Metascraper Metadata Extractor turns a known list of public page URLs into one structured dataset record per page. It returns normalized title, description, author, publication date, image, logo, publisher, language, and URL alongside Open Graph, Twitter card, canonical, favicon, JSON-LD, and raw head-tag provenance for teams building preview, content, and SEO workflows.

Best fit and focused standalone workflow

Choose this Actor when you already have page URLs and want page-authored metadata in a compact, API-ready dataset. It works as a focused standalone workflow for link-preview QA, known-URL content enrichment, and canonical-field reviews. Start with a specific page list, then pass the returned source URL, canonical URL, title, publisher, and metadata image into your CMS, QA sheet, or content pipeline for the next enrichment or review step.

Workflow story: from URLs to a preview-ready record

A content operations team starts with a spreadsheet of article URLs. Next, it supplies those URLs to this Actor and receives one normalized record per page with the author, date, canonical URL, metadata image, and source-tag provenance. The team then compares the returned fields, feeds clean rows into its publishing workflow, and schedules a later run when it wants a fresh collection.

Input

Pass urls as a plain array or startUrls in the Apify request-list format. The Actor accepts up to 100 distinct HTTP(S) URLs and processes 25 by default. includeRawMetaTags preserves grouped page-head metadata; includeJsonLd keeps valid JSON-LD blocks for structured-data context.

{
  "urls": ["https://example.com"],
  "includeRawMetaTags": true,
  "includeJsonLd": true,
  "maxResults": 25
}

Output

Each page produces one dataset item. metadata contains the normalized Metascraper result, while openGraph, twitterCard, and rawMetaTags retain page-authored provenance.

Field	Meaning
metadata.title	Normalized page title selected by Metascraper rule bundles.
canonicalUrl	Declared canonical link resolved against the final page URL.
openGraph	Page-authored og:* tags, including repeat values where present.
twitterCard	Page-authored twitter:* tags for preview debugging.
jsonLd	Valid page JSON-LD blocks, capped at 20 per page.
warnings	Page-level provenance notes, such as malformed JSON-LD blocks.

{
  "sourceUrl": "https://example.com/",
  "finalUrl": "https://example.com/",
  "statusCode": 200,
  "metadata": {
    "title": "Example Domain",
    "description": null,
    "author": null,
    "publishedAt": null,
    "image": null,
    "logo": null,
    "publisher": null,
    "language": "en",
    "url": "https://example.com/"
  },
  "canonicalUrl": null,
  "openGraph": {},
  "twitterCard": {},
  "warnings": []
}

Pricing

The Actor uses Pay per Event with Apify platform usage included. The metadata-extracted event is $0.004 for each successful persisted metadata record and Actor start is $0.00005. A 25-page collection has an event-cost cap of $0.10005 before platform usage. Review the live Pricing tab before a production workflow because it is the current source of truth for event and platform-usage charges.

Run through the API

export APIFY_TOKEN="your-apify-token"
curl "https://api.apify.com/v2/acts/khadinakbar~metascraper-metadata-extractor/run-sync-get-dataset-items?token=${APIFY_TOKEN}" \
  -X POST \
  -H "Content-Type: application/json" \
  -d '{"urls":["https://example.com"],"includeRawMetaTags":true,"includeJsonLd":true,"maxResults":1}'

AI agent prompt

Use Metascraper Metadata Extractor for these supplied public page URLs. Return the title, canonical URL, normalized image, publisher, Open Graph tags, JSON-LD, and warnings. Read the dataset results after completion, retain source URL provenance, report the terminal outcome, and keep the collection within the stated per-record cost and page scope.

The terminal OUTPUT and RUN_SUMMARY records provide outcome and dataset-readback context. A supplied page list is the full scope of a run, so agents can control collection size and event cost with maxResults.

Best-results guidance

Provide canonical public pages with complete HTML head sections and use the final URL, status code, warnings, and source-tag groups together when reviewing a record. For recurring checks, schedule the same bounded URL list and compare the fresh dataset against the prior collection in your own workflow.

Builder's note

I built this Actor around Metascraper’s official, composable rule bundles because a normalized result alone can hide the field source that a QA or ingestion team needs to inspect. My design keeps both layers: a concise normalized record for downstream use and raw tag provenance for an informed follow-up decision.

Responsible use

Use this Actor for public pages your organization is authorized to access, with applicable laws, site terms, and internal data policies guiding each collection.

Call over HTTP

Use the Actor's actual URL in the Endpoints tab with an Apify bearer token. GET /metadata?url=https%3A%2F%2Fexample.com supports one public URL and scalar options. POST /metadata accepts the existing input fields and optional maxCostUsd. It returns records plus summary directly; normal batch input and exports remain available.

The unchanged charge is $0.004 per page with usable successfully extracted metadata. Existing batch startup-event pricing is unchanged. Platform usage is included under the current creator-paid setting; no free-user quota, paid feature or price was changed. Readiness and rejected-input/insufficient-budget requests do no source work and charge no result events. A metadata fetch/error/challenge without usable metadata is not a billable empty result. An uptime DOWN observation is useful monitor output and remains billable. No new per-request startup event is manually added.

maxCostUsd is checked before source work for every planned observation. It bounds this request's result events and does not enforce a Console-wide allowance. Completed result events are charged before delivery with the platform request ID. If charging is uncertain, the response is withheld and should not be retried automatically. Useful partial responses use HTTP206 with complete records and explicit counts.

HTTP responses do not create a caller-specific run, Dataset export or history entry. Caller URLs, content, results and proxy credentials are not stored in shared Dataset/KV or general logs. Only aggregate counters are retained. Use normal runs for persistent exports and large jobs.

The service accepts2 active requests per instance, at most1MiB of JSON input,2MiB of cumulative decoded source response bytes across attempts and targets, and2MiB of complete output. It rejects oversized output before result billing instead of truncating fields. Work has a180-second soft admission window and240-second overall deadline; full long/bulk workflows remain on batch. Existing per-run caps and all normalized fields, raw head tag/JSON-LD and proxy options remain.

Only public HTTP(S) targets without URL credentials are accepted. Socket-bound DNS/IP checks reject private/reserved destinations and unsafe redirect targets. Managed Apify Proxy settings and caller-supplied public proxy configurations are isolated per request; proxies can add platform usage cost. No customer login, cookies, browser rendering, custom request body or mutating target method is offered. Missing or unusable proxy configuration returns a clear error on HTTP instead of silently changing the requested network route.