Website Screenshot: Viewport & Full-Page Captures

Urlwright. Little tools for the web.

Clean screenshots of public pages. Viewport or full page, PNG or JPEG. One dataset row per URL, and the image in the key-value store.

v1 opens each page in a fresh headless Chromium window. It does not log in, fill forms, or keep a cookie jar. A site that needs a session will not show the logged-in page.

What it does

1. You pass page URLs (startUrls) or a dataset of URLs. startUrls accepts plain strings and Apify request objects { "url": "…" } (mixed lists are fine). Bare domains are normalised to https://.
2. Social hosts are rejected before any fetch: Facebook, Instagram, LinkedIn, X (Twitter), TikTok, Threads, Pinterest, Snapchat, Reddit, and the short domains fb.com, fb.me, lnkd.in, t.co, pin.it, redd.it. Subdomains of those hosts are rejected too.
3. Each remaining URL is checked for a private, link-local or metadata address. A blocked URL is written as ssrf-blocked and is not opened.
4. When respectRobotsTxt is true (the default), the actor fetches that origin's robots.txt, follows up to five redirects, and caches the file for the run. A matching Disallow, a 5xx, or a failed fetch is written as robots-blocked and is not opened. A 4xx or an empty file allows the page. If the robots URL itself is a private address, the row is ssrf-blocked. If a redirect of that file is a private address, the row is robots-blocked.
5. Each remaining URL is opened in headless Chromium and captured. The default is the viewport, 1280×720, not the full scroll height. fullPage: true captures the scrollable page.
6. The image is stored under screenshotKey (screenshot-0001.png or screenshot-0001.jpg). The dataset row carries status, HTTP status, size and that key.
7. A successful file charges the event screenshot-captured once. A blocked URL, a failed capture and a duplicate URL in the same run are free.

One URL that fails does not fail the run. RUN_SUMMARY is always written to the key-value store, including when the input is rejected.

Who it is for

• A visual record of a public page you can open without an account.
• A fixed desktop viewport (1280×720 by default), or the full scroll height when you turn fullPage on.
• A list of URLs, typed or taken from another dataset, turned into PNG or JPEG files plus one row each.

Logged-in dashboards and the social hosts above are out of scope for v1.

Input

The console prefills a small run (https://urlwright.com/, 5 screenshots). If you omit maxScreenshots, the default is 50. The hard cap in the input schema is 500.

{
  "startUrls": ["https://urlwright.com/"],
  "datasetId": "",
  "datasetField": "url",
  "fullPage": false,
  "viewportWidth": 1280,
  "viewportHeight": 720,
  "format": "png",
  "quality": 85,
  "waitUntil": "networkidle",
  "waitSecs": 0,
  "hideCookieBanners": false,
  "timeoutSecs": 60,
  "maxScreenshots": 5,
  "respectRobotsTxt": true
}

fullPage defaults to false. viewportWidth is 320–3840 (default 1280). viewportHeight is 240–2160 (default 720). With fullPage true the height still sets the initial window; the image is the scrollable page.

format is png or jpeg. quality (1–100, default 85) applies only to JPEG. waitUntil is load, domcontentloaded or networkidle (default networkidle). timeoutSecs is 5–120 (default 60) for navigation and the screenshot.

waitSecs is an integer from 0 to 30 (default 0). After the page reaches waitUntil, the actor waits that many seconds (page.waitForTimeout), then takes the screenshot. 0 skips the wait. The wait is on top of timeoutSecs.

hideCookieBanners defaults to false. When it is true, the actor hides common cookie consent banners (including Cookiebot and Usercentrics) with CSS, calls the Usercentrics consent API when the page provides it, and clicks one common accept or reject button. If that dismiss fails, the capture still runs. It does not log in, fill forms, or keep a cookie jar. It does not change the price. For the settings that keep a hero clear, see Clean website screenshots without cookie banners.

datasetId plus datasetField (default url) reads extra targets. If the named field is missing on a row, the actor tries url, then finalUrl, then domain, then inputUrl. A bad dataset row is skipped. Duplicate URLs in one run are collapsed (trailing slash ignored).

respectRobotsTxt defaults to true. A disallowed path becomes a free robots-blocked row. Set it to false to capture that path anyway.

There is no username, password, cookie, proxy or storage-state field.

Output

One dataset. Every field is present. Unused fields are null.

A captured page:

{
  "schemaVersion": 1,
  "inputUrl": "https://urlwright.com/",
  "url": "https://urlwright.com/",
  "domain": "urlwright.com",
  "status": "ok",
  "httpStatus": 200,
  "finalUrl": "https://urlwright.com/",
  "error": null,
  "fullPage": false,
  "viewportWidth": 1280,
  "viewportHeight": 720,
  "format": "png",
  "bytes": 32346,
  "screenshotKey": "screenshot-0001.png",
  "charged": true,
  "chargedEvents": { "screenshot-captured": 1 },
  "fetchedAt": "2026-10-26T08:00:00.000Z"
}

domain is the hostname without a leading www.. bytes is the file size. screenshotKey is screenshot- plus a four-digit index, then .png or .jpg. The bytes live in the key-value store under that key (image/png or image/jpeg). View: Screenshots.

status is one of:

Status	Meaning	Charged
ok	Image stored	yes, screenshot-captured × 1
social-blocked	Host on the social list; no fetch	no
ssrf-blocked	Private, link-local or metadata address	no
robots-blocked	robots.txt disallows the path, or the file was unreachable (5xx or network)	no
error	Navigation or screenshot failed	no
invalid-input	The input was rejected; one row, then the run stops	no

An ok row also has screenshotUrl: a signed Apify key-value record URL for that file (…/records/<screenshotKey>?signature=…). Under account Restricted General resource access this link opens anonymously without an API token; it stays valid while the record exists (and until the store’s URL-signing key rotates). screenshotKey is still the KV key. deviceScaleFactor is echoed on every row.

A blocked or failed row has charged: false, screenshotKey: null, screenshotUrl: null, bytes: null, and chargedEvents with zero counts for all capture events. error is social-blocked:<host>, ssrf-blocked, robots-blocked, or the capture message.

RUN_SUMMARY counts targets, ok, failed, socialBlocked, ssrfBlocked, robotsBlocked, and chargedEvents, and echoes deviceScaleFactor, waitSecs and hideCookieBanners. stoppedReason is completed, no_targets, invalid_input or aborted. On invalid input the summary also has fatalError.

A URL already captured in the same run is not captured or billed again when that run is migrated or resurrected.

Pricing

Pay per event. Urlwright pays platform compute. One charge per successful image file; the event name depends on deviceScaleFactor. There is no start event. Social blocks, SSRF blocks, robots blocks, failed captures and duplicates are free. From 21 Nov 2026, a run with at least one paid result is topped up to $0.01 via minimum-fee ($0.0001 units on every tier); empty or blocked runs stay free.

Event	Scale	FREE	BRONZE	SILVER	GOLD
screenshot-captured	1×	$0.016	$0.015	$0.0145	$0.014
screenshot-captured-2x	2×	$0.024	$0.0225	$0.02175	$0.021
screenshot-captured-3x	3×	$0.032	$0.030	$0.029	$0.028
screenshot-captured-4x	4×	$0.040	$0.0375	$0.03625	$0.035
minimum-fee	—	$0.0001	$0.0001	$0.0001	$0.0001

PLATINUM and DIAMOND match GOLD. Minimum max total charge per run: $0.05 (so one 4× capture fits). Prices are live on the private actor (GET-verified). Margin evidence: docs/margin-evidence.md.

Limits and runtime

Suggested run options: build latest, 2048 MB memory, timeout long enough for your URL count. Default memory is 2048 MB. The maximum is 4096 MB. A single public page on urlwright.com finished navigation and screenshot in about one second at 2048 MB (viewport and full page). 4096 MB did not improve that page. Heavier pages can need the higher memory.

Each URL starts its own headless Chromium. The browser user agent is UrlwrightBot/1.0 (website-screenshot; +https://urlwright.com/bot). JavaScript runs. Cookies and storage state are not loaded. The start URL is rejected when it resolves to a private, link-local or metadata address, or when DNS fails. That row is free.

respectRobotsTxt defaults to true. Before a capture, the actor fetches /robots.txt on that origin with the Urlwright user agent, follows up to five redirects, and caches the file for the rest of the run. A Disallow that matches the page, a 5xx, or a network failure skips the URL: status robots-blocked, no image, no charge. A 404, any other 4xx, or an empty file allows the capture. The robots URL goes through the same private-address guard as the page. If that URL itself is blocked, the row is ssrf-blocked. If a redirect of robots.txt is blocked, the file counts as unreachable and the row is robots-blocked. Set respectRobotsTxt to false to skip this check.

v1 does not emulate a phone model, switch a colour scheme, or compare two screenshots. hideCookieBanners (default false) is a best-effort banner hide, not a login. waitSecs (0 to 30, default 0) pauses after waitUntil and before the screenshot. The image is the page as served to the viewport you set (or the full scroll height when fullPage is true), at deviceScaleFactor 1–4 (default 1). Higher scale means more device pixels for the same CSS viewport — sharper bitmaps, larger files, higher event price.

The schema allows at most 500 screenshots. The console prefill is 5. The default when the field is omitted is 50. Each page waits at most timeoutSecs (default 60).

FAQ

Does it log in or keep cookies? No. No username, password, cookie jar or saved browser state. You get the page an anonymous browser gets.

Why was a social URL not captured? Facebook, Instagram, LinkedIn, X, TikTok, Threads, Pinterest, Snapchat, Reddit and their short domains are rejected before any request. The row is social-blocked and is free.

Does it respect robots.txt? Yes, by default. It downloads robots.txt for the page origin and skips a path UrlwrightBot is disallowed from. That row is robots-blocked and free. A missing file (404) allows the capture. A server error or a network failure while reading robots.txt also skips the page. Set respectRobotsTxt to false to capture anyway.

Viewport or full page? The default is the 1280×720 viewport (fullPage: false). fullPage: true captures the scrollable page. Both use the same event when the file is stored.

PNG or JPEG? format is png (default) or jpeg. quality applies only to JPEG (default 85). The key ends in .png or .jpg.

Will one bad URL fail the run? No. Other URLs continue. Rejected input writes one invalid-input row and stops. RUN_SUMMARY is always written.

Can it hide a cookie banner? Yes, when hideCookieBanners is true. It hides common consent banners (including Cookiebot and Usercentrics) with CSS, calls the Usercentrics consent API when the page provides it, and clicks one common accept or reject button. If that fails, the screenshot still runs. It does not log in and it does not keep a cookie jar. There is no phone preset, no dark-mode switch and no before/after comparison. Set viewportWidth / viewportHeight for the window size, and deviceScaleFactor 2–4 for retina-style bitmaps (@2x / @3x / @4x).

Can it wait for a late hero? Yes. waitSecs (0 to 30, default 0) waits that many seconds after the page is ready and before the screenshot. Use a few seconds on marketing sites that stay blank at domcontentloaded. The same capture event is charged. A longer wait uses more compute time, not a new event.

What does a screenshot cost? One successful file is one event. At 1× (default) that is screenshot-captured (FREE $0.016 / GOLD $0.014). At 2× / 3× / 4× the events are screenshot-captured-2x / -3x / -4x with higher list prices. Blocked and failed URLs are free. See the Pricing table.

What is device scale / @2x? deviceScaleFactor is Playwright’s device pixel ratio. 1 keeps CSS pixels = image pixels. 2 writes roughly four times as many pixels for the same viewport (retina-style). It does not make the site “better”; it only increases bitmap sharpness and file size. Prefer 1× unless you need a sharp asset for high-DPI displays.

Where is the image URL? On ok rows, screenshotUrl is a signed Apify KV record URL (?signature=). You can open it in a browser or curl without an API token even when the account uses Restricted General resource access (Anyone-with-ID would also work with the unsigned form, but signed is what we emit). The link is permanent for that record for as long as the file remains in the store. You can also download via screenshotKey in the run’s default key-value store with a token.

No warranty

The output depends on the sites you open. Pages change, block automated browsers, or look different once a person is logged in. A screenshot is not proof of how a page appeared to every visitor, and it is not legal advice. No warranty is given that the capture matches a particular browser, or that the file remains available after the run storage expires.