Screenshot APIs you can pay per call

What a screenshot API returns

A screenshot API loads a web page in a browser it controls and returns an image of the rendered result. The input is usually a URL, sometimes raw HTML, and the output is a PNG, JPEG or WebP file, often encoded as base64 inside a JSON response. The image shows layout, fonts, images and visible text after scripts have run. That makes it useful when the HTML alone cannot answer the question, for example whether a banner covers the checkout button or a chart rendered at all.

Agents use captures in three broad ways. A deploy check compares the new page with an expected layout. A vision model reads the image when the page content is drawn on a canvas or assembled by scripts. A record of what a page displayed at a given moment supports later review. Each use asks something different of the capture: a deploy check needs a stable viewport, a vision model needs legible text at a manageable file size, and a record needs the URL, time and settings kept with the file.

A screenshot is a rendering by one browser, at one size, from one network location. Personalization, geography, cookie consent, A/B tests and bot protection can all change what appears. A capture that succeeds technically can still show a consent wall, an error page or a blank region where a slow script had not finished. Treat the image as the provider's rendering of the request you sent, and check it before relying on it.

Page text and screenshots are related products with different outputs. A reader route returns words and links; a capture route returns pixels. When a workflow needs both, call both and keep them linked, rather than assuming one stands in for the other.

Captures, PDF exports and archive bundles

The comparison treats an endpoint as a capture when its documented output is an image of the rendered page. A PDF export of a page is a neighboring product: it keeps a printable layout and selectable text, but its pagination and print styles can differ from what a screen shows. A capture bundle adds rendered HTML, metadata or a signed receipt to the image. A service enters the comparison only with a priced image capture; a PDF or bundle quote never supplies the row price on its own.

App-store listings that return screenshot URLs, browser sessions that include screenshots as actions and page readers that return text stay outside the capture type. They answer other questions, and their prices describe other units of work.

Capture options vary by endpoint. In the matched endpoint descriptions:

• Full-page capture is documented by Hugen Visual API, Agent Web Access, agentsvc.io, agdata, ISHOB, true402, Crypto Technicals & Web Utilities and ArkBrowser API.
• Capture of an element or an exact region is documented by Hugen Visual API, agdata and ArkBrowser API.
• Device emulation is documented by Hugen Visual API.

An option missing from a description is not proof that the route lacks it; it means the documented contract does not promise it.

Rendering engines are named by the providers. A Chromium renderer is named in the matched endpoint descriptions of ISHOB, true402, Crypto Technicals & Web Utilities and ArkBrowser API.

Playwright is named by agentsvc.io. The engine affects font rendering, supported web features and how scripts behave, but a named engine is a provider statement, not an observation of the version that served your request.

Each price is the cheapest endpoint of that service for that kind of data: the same number wherever this page quotes it.

Compare 10 screenshot APIs

Each row is priced on its cheapest priced capture endpoint, with that path kept in the row. Rows are ordered by observed paying wallets over 30 days; the order describes usage, not image quality.

$/CALL is the price of the endpoint named under it: the cheapest rendered page captured as an image endpoint of the service or, for a service without one, its cheapest screenshot endpoint. Never the cheapest tool of a gateway. Other kinds of data have their own price in the tables below. SCREENSHOT ENDPOINTS counts the screenshot endpoints out of all the endpoints the service lists. BUYERS 30D counts distinct wallets that paid the whole service on-chain in 30 days, not only its screenshot endpoints. "Generalist" marks a gateway that sells more than three unrelated kinds of tools: it is never called the cheapest.

How to choose a screenshot API

Fix the capture settings first

Decide the viewport width and height, whether you need the full scrolling page, and which output format the next step accepts. Full-page captures of long pages can reach the provider's height limit or produce files too large for a vision model. A width of 1280 pixels and a width of 390 pixels show different layouts on responsive sites, so pick the sizes your users actually see.

Decide when the page is ready

Pages finish loading at different moments. A route with a wait setting, a selector to wait for, or browser steps before capture gives you control over that moment. Without it, a capture can arrive before web fonts, lazy images or client-side data. Include slow pages and pages with consent dialogs in your test set.

Check access rules and failure billing

Access and billing terms differ by route. A provider may respect robots.txt, decline to pass bot protection or state that failed and blank captures are not charged. These terms decide what an empty result means and whether it costs money. Read them for the exact route, and record the reason when a capture fails rather than retrying blindly.

Plan for storage and privacy

A base64 image inside a JSON response must be decoded and stored by you. A route that uploads the file to provider storage or keeps downloads for a period moves a copy of the page onto someone else's infrastructure. Avoid capturing pages that show personal data unless the provider's retention terms fit your obligations.

Measure on your own pages

The median monitored uptime of the listed services is 100%. Monitoring checks the payment endpoint, not how long a heavy page takes to render, so time a representative set of your own URLs at the concurrency you need.

Which endpoint fits which job

Start from the file your workflow keeps. The candidates below come from the same endpoint classification as the table, and each quote keeps its path.

Capture a page for visual review

Fix the viewport, wait condition and capture area before comparing providers. A blank or cookie-covered image is a failed result even when the request returns a successful status.

Candidates today, most buyers first: Hugen Visual API ($0.02/call, 312 buyers) · Agent Web Access ($0.005/call, 111 buyers) · agentsvc.io ($0.009/call, generalist, 41 buyers) · agdata ($0.01719/call, generalist, 19 buyers) · ISHOB ($0.001/call, generalist, 16 buyers) · true402 ($0.004/call, generalist, 7 buyers) · Crypto Technicals & Web Utilities ($0.01/call, generalist, 3 buyers) · x402.eleeth.com ($0.05/call, generalist, 1 buyer) · ArkBrowser API ($0.015/call) · Web Capture ($0.1/call)

Keep a printable PDF of a page

Check page size, margins and whether scripts run before printing. A PDF export preserves layout for reading; it does not certify when or how the page was served.

Candidates today, most buyers first: Agent Web Access ($0.005/call, 111 buyers) · agentsvc.io ($0.0045/call, generalist, 41 buyers) · agdata ($0.01719/call, generalist, 19 buyers) · Web Capture ($0.1/call)

Preserve a capture with rendered HTML

Keep the image, the rendered HTML and any receipt together with the requested URL and capture time. Read the provider's limits on redirects, third-party assets and retention.

Candidates today, most buyers first: Web Capture ($0.1/call)

What the main providers describe

These notes repeat what each provider documents for its capture route. Options, limits and storage terms are the providers' statements; prices come from the matching endpoint type rather than from figures in their descriptions.

Hugen Visual API

Hugen Visual API describes a single capture route, GET /visual/screenshot, at $0.02. The provider documents JPEG or PNG output with page title, file size and render time, plus full-page scroll capture, CSS selector targeting, device presets and dark mode. It also advertises ad and cookie-banner removal. Removing overlays changes what the image shows, so keep that setting in the capture record when the image is evidence of what a visitor saw.

1 screenshot endpoint of 1 · $0.02/call (GET /visual/screenshot) · Base/SOL · uptime 30d 99.9% · 1459 ms · 312 buyers in 30 days

Agent Web Access

Agent Web Access lists GET /screenshot at $0.005 and a separate PDF render at $0.005. The provider describes a real browser at a chosen viewport width, with viewport-only or full-page output returned as base64 PNG. It positions the capture for deploy checks, layout review and vision-model input. The same host lists page text and CSS-selector extraction; those routes return text, so they cannot stand in for the image when a reviewer needs to see the page.

2 screenshot endpoints of 4 · $0.005/call (GET /screenshot) · Base · uptime 30d 99.6% · 209 ms · 111 buyers in 30 days

agentsvc.io

agentsvc.io lists POST /screenshot at $0.009 and POST /html-to-pdf at $0.0045. The provider names Playwright with full JavaScript rendering, a default width and height, a full-page flag and a wait setting. Its PDF route accepts a URL or a raw HTML string, with format, orientation and margins. The host also sells scraping, OCR and token checks, so record the path you called rather than the package name.

2 screenshot endpoints of 50 · $0.009/call (POST /screenshot) · ARB/Base/POL/SOL · uptime 30d 100% · 56 ms · 41 buyers in 30 days

ISHOB

ISHOB describes POST /v1/screenshot at $0.001, rendered in headless Chromium through Cloudflare Browser Rendering. The provider documents PNG, JPEG or WebP output, a viewport up to 1920 by 1920 and full-page capture up to 8,000 pixels. It states that robots.txt is respected, page requests are SSRF-guarded and nothing is stored. A page that disallows automated access can therefore produce no image, which is a different outcome from a rendering failure.

1 screenshot endpoint of 34 · $0.001/call (POST /v1/screenshot) · Base · uptime 30d 100% · 173 ms · 16 buyers in 30 days

Web Capture

Web Capture sells one bundle, POST /v1/captures, at $0.1: a browser screenshot, rendered HTML, metadata, a PDF report and a signed artifact receipt. The provider limits its own claims. Time comes from the service clock, there is no independent timestamp or completeness guarantee, and assets or redirects on other hosts are excluded. Treat the receipt as integrity evidence for the files produced, not as a legal certification of the page.

1 screenshot endpoint of 1 · $0.1/call (POST /v1/captures) · Base · uptime 30d 97.9% · 1769 ms · 0 buyers in 30 days

What captures cost at volume

The median selected capture costs $0.0125 per call, so 100,000 captures total $1,250.00 at that rate. The table repeats the calculation for the median, the cheapest specialized service and the providers described above.

Each service at the price of its main screenshot endpoint, the one named in the table above.

A capture call is one rendering under one set of options. A monitoring job that captures ten pages at two widths every hour makes 480 calls a day, so the per-call quote multiplies quickly. Full-page captures, PDF exports and bundles can carry different prices on the same host. Retries after a timeout, captures that show a consent wall and pages that need a second attempt with a longer wait all add calls.

Set a budget per page and per run, and stop when repeated captures fail for the same reason. A cheaper route that needs two attempts per usable image can cost more than a dearer route that succeeds once. Keep the quoted path with each charge so the bill can be reconciled with the images you kept.

Inspect a capture request and its payment terms

For a capture, Hugen Visual API documents GET /visual/screenshot with a reference quote of $0.02. Read its current request schema for the URL, format, full-page, selector and device options before paying, and check that the response fields match what your workflow stores.

The listed method and description establish the advertised contract. This explanation does not reproduce an HTTP 402 challenge or certify a paid response. Request the challenge for the exact call you intend to make and compare resource, method, network, asset and amount before authorizing payment. A challenge for a PDF route on the same host does not authorize a screenshot.

How this comparison is built

The comparison starts from services that are online and payment-ready, with a known price, a description of at least 80 characters and measured 30-day uptime of at least 90%. Endpoint selection then reads each service's documented routes. A service enters the table only when it lists a priced endpoint whose described output is a rendered page image. PDF exports and capture bundles are classified separately and shown as additional types for services that also capture images.

Provider descriptions and endpoint lists were read on October 22, 2026, from the public listing. Output formats, limits and storage terms quoted here are provider claims, not tests of image quality or rendering accuracy. Buyer counts are distinct paying wallets observed over 30 days. Monitoring records uptime and response time of the payment endpoint; it does not render pages or compare images.

Prices are the reference amounts listed for each endpoint. A route whose live amount depends on session state or options can quote differently for your request, so the reference price is a starting point rather than a promise.

Questions

How to pay

Each service answers an unpaid call with HTTP 402 and its price; your agent signs a USDC payment on the listed network and retries. No API key, no account, no subscription. See what x402 is.