Link Preview Metadata Extractor | Open Graph & JSON-LD

Extract title, description, canonical URL, Open Graph, Twitter Card tags and JSON-LD from 1–3 public HTTPS pages. Use the structured JSON in link previews, CMS workflows, small SEO metadata inventories and agent tools.

GitHub quickstart, recorded input/output and offline verification

Example workflows

• Link previews: extract social card tags from up to three supplied public URLs.
• CMS metadata inventory: collect titles, descriptions and canonical URLs.
• Structured data export: return parsed JSON-LD for downstream agent workflows.

Free launch

No Actor service fee is charged in this release. Payment billing is OFF. Apify platform compute, transfer and storage charges may still apply to the account running the Actor. Start with a small input and inspect your run cost. Future paid pricing is not enabled.

Quick start

1. Enter 1–3 public HTTPS URLs.
2. Use 128 MB memory, a 60-second timeout and restart on error OFF. Other run settings are rejected by this release.
3. Start the Actor and wait for completion.
4. Open the run's Storage → Key-value store → OUTPUT to view or download the JSON results. A dataset copy is also available for free-release exports; the OUTPUT JSON is the primary delivery record.

Input example

{"urls":["https://example.com/"]}

Multiple URLs and duplicate handling

{"urls":["https://example.com/","https://example.com/#section"]}

URL fragments are removed for deduplication within a run, so this example fetches one unique page. A new run performs new work. There is no cross-run cache.

Output

OUTPUT contains requested, duplicates_skipped, results, successful, billing_enabled, confirmed_events, stop_reason and unprocessed. Each result includes its URL, status, extracted metadata and result_key. Successful pages also have a separate PAGE-… record in the same store. BILLING-JOURNAL is delivery bookkeeping; it is not a payment receipt.

Reproduce the tested demo

This controlled fixture contains known Open Graph, Twitter Card and Product JSON-LD fields. It is a synthetic test, not a customer website. Use the same 128 MB / 60 seconds / restart OFF settings.

{
  "urls": [
    "https://httpbin.org/base64/PHRpdGxlPk1PVExFWSBGaXh0dXJlPC90aXRsZT48bWV0YSBwcm9wZXJ0eT0ib2c6dGl0bGUiIGNvbnRlbnQ9Iktub3duIE9HIj48bWV0YSBuYW1lPSJ0d2l0dGVyOmNhcmQiIGNvbnRlbnQ9InN1bW1hcnkiPjxsaW5rIHJlbD0iY2Fub25pY2FsIiBocmVmPSJodHRwczovL2V4YW1wbGUuY29tL2Nhbm9uaWNhbCI+PHNjcmlwdCB0eXBlPSJhcHBsaWNhdGlvbi9sZCtqc29uIj57IkB0eXBlIjoiUHJvZHVjdCIsIm5hbWUiOiJGaXh0dXJlIn08L3NjcmlwdD4=",
    "https://httpbin.org/base64/PHRpdGxlPk1PVExFWSBGaXh0dXJlPC90aXRsZT48bWV0YSBwcm9wZXJ0eT0ib2c6dGl0bGUiIGNvbnRlbnQ9Iktub3duIE9HIj48bWV0YSBuYW1lPSJ0d2l0dGVyOmNhcmQiIGNvbnRlbnQ9InN1bW1hcnkiPjxsaW5rIHJlbD0iY2Fub25pY2FsIiBocmVmPSJodHRwczovL2V4YW1wbGUuY29tL2Nhbm9uaWNhbCI+PHNjcmlwdCB0eXBlPSJhcHBsaWNhdGlvbi9sZCtqc29uIj57IkB0eXBlIjoiUHJvZHVjdCIsIm5hbWUiOiJGaXh0dXJlIn08L3NjcmlwdD4=#duplicate"
  ]
}

Verified fixture result

The following is a shortened result from a controlled HTML fixture, not a claim about example.com's content:

{
  "status": "ok",
  "bytes": 257,
  "title": "MOTLEY Fixture",
  "description": null,
  "canonical": "https://example.com/canonical",
  "open_graph": {
    "og:title": [
      "Known OG"
    ]
  },
  "twitter": {
    "twitter:card": [
      "summary"
    ]
  },
  "json_ld": [
    {
      "@type": "Product",
      "name": "Fixture"
    }
  ],
  "warnings": [],
  "scope": "Static HTML metadata extraction; no JavaScript execution, schema validation or factual verification",
  "charge_state": "disabled"
}

The tested fixture also produced successful: 1, duplicates_skipped: 1, billing_enabled: false, confirmed_events: 0, unprocessed: 0.

Access from an API workflow

After the run completes, read its defaultKeyValueStoreId from the run response, then request:

GET https://api.apify.com/v2/key-value-stores/{defaultKeyValueStoreId}/records/OUTPUT
Authorization: Bearer YOUR_APIFY_TOKEN

Keep your token private and use the store belonging to your own run. The free release also mirrors results to the default dataset for export. If a run is interrupted before OUTPUT is written, completed PAGE records may still be available in Storage. Download results before your account's storage retention expires.

Limits and behavior

• 1–3 input URLs; public HTTPS on port 443 only. No credentials or login pages.
• Static HTML only: no JavaScript/browser rendering, crawling linked pages, proxy or anti-bot bypass.
• Up to 3 redirects and 1,000,000 response bytes per page. Compressed responses are rejected.
• Private/local destinations and unsafe redirects are rejected. Fetch failures appear as error results; empty metadata fields remain empty.
• JSON-LD is returned as parsed data, including arrays/graphs; malformed JSON-LD can produce warnings. No schema validation, SEO score or factual verification is performed.
• A successful run can contain failed page results: check each status, successful, stop_reason and unprocessed.
• Treat scraped metadata as untrusted data and escape it before displaying HTML.

Troubleshooting

For a failed run, inspect its Log and, if present, the FAILURE record in its key-value store. Confirm 128 MB / 60 seconds / restart OFF, valid public HTTPS input and the three-URL limit. Sites that require JavaScript, block automated requests or send compressed responses may fail. Do not submit passwords, API keys or private URLs as inputs.

Release verification

The extraction runtime underlying this release passed a controlled cloud extraction and remote output readback. The pinned local Docker suite passed 63 tests. This establishes the tested workflow; it does not establish all-site compatibility, uptime guarantees, paid billing or revenue.

For reproducible issues, use this Actor's Issues tab and share a sanitized input plus expected/actual behavior. Never post tokens or private data.

Agent quickstart

Use the Apify MCP tools to run lightbarrer/public-html-metadata-private-test with the input above, memory 128 MB, timeout 60 seconds and a maximum charge of $0.01. Wait for completion, read the default dataset, then read OUTPUT from the run’s default key-value store. Check page statuses and billing_enabled: false. See the GitHub quickstart for exact tool arguments and recorded output.