The screenshot API that understands the page.

Capture pixel-accurate webpages and turn them into structured visual context agents can inspect. Full-page screenshots, browser interactions, deterministic regions, visual diffs, monitoring and a screenshot-grounded Visual Style Guide, through one API.

Start free Read the docs

100 credits a month, free. No card required.

RequestcURL
curl "https://api.scrinly.com/render/screenshot" \
  -X POST \
  -H "Authorization: Bearer $SCRINLY_API_KEY" \
  -H "Content-Type: application/json" \
  --data-raw '{"url":"https://tailwindcss.com/","fullPage":true,"regions":true}' \
  --fail-with-body  \
  --silent --show-error

# One browser session creates the screenshot
# and its deterministic region crops, all tied
# to the same image hash.
Responseapplication/json
{
  "success": true,
  "type": "screenshot",
  "media": {
    "stored": true,
    "key": "homepage/tailwind/59305565...jpg",
    "url": "https://.../59305565...jpg",
    "contentType": "image/jpeg"
  },
  "regions": {
    "version": "regions-v1",
    "screenshot": {"width":1600,"height":11593,"sha256":"59305565..."},
    "items": [{"id":"region-0","kind":"navigation", …}],
    "coverage": {"ratio":1, "truncated":false, "omittedAssets":0}
  },
  "credits": {"charged":3,"refunded":0,"net":3}
}

One screenshot.
Three layers of context.

Capture the page as users saw it, break long pages into context an agent can inspect, and add a visual style guide only when the workflow needs deeper understanding.

CAPTURE — 01

Pixel-accurate capture

Give agents and automations a faithful record of the exact page state across viewports, devices, themes, and interactions.

REGIONS — 02

Deterministic regions

Make long pages readable to vision models by returning ordered sections they can inspect without shrinking the entire page into one image.

EVIDENCE — 03

Compare, monitor, describe

Catch visual regressions, monitor important pages, and turn the same screenshot evidence into a reusable visual reference.

One screenshot.
Traceable results.

Screenshots, regions, style guides, and visual diffs all point back to the same captured image. Hashes and metadata make it easy to check where each result came from.

POST /render/screenshot
RequestcURL
Responseapplication/json

Direct deliveryScrinly B2 or your B2, S3 or R2 bucket returns the provider URL directly. Content-addressedDifferent encoded bytes receive different keys; identical bytes can safely reuse one. Partial by designA region or Style Guide failure retains the screenshot and refunds that increment.

Capture the page,
exactly as it appeared

Capture the exact visual state your workflow needs across devices, themes, and post-interaction states—without operating a browser or maintaining capture infrastructure.

One request per device preset. Each capture returns its own result, with its own stored media and screenshot hash evidence.

tailwindcss.com captured at iPhone 15 size
iphone-15393 × 852
tailwindcss.com captured at iPad Air size
ipad-air820 × 1180
tailwindcss.com captured at MacBook Pro 14 size
macbook-pro-141512 × 982

Faithful by defaultCleanup and blocking controls run only when you enable them. Act before captureClick, scroll, hide selectors or wait for the network to settle. Store or streamReturn image bytes directly or store them in Scrinly B2 or your own bucket.

Know what changed,
not just that it changed

Catch visual regressions before users do. Compare releases, monitor critical pages, and give agents a focused record of what changed, where it changed, and whether it needs attention.

OverlayReview
Annotated overlay showing changes in a full-page Tailwind CSS capture
HeatmapDensity
Contextual heatmap showing change density in a full-page Tailwind CSS capture
MaskMachine output
Binary mask showing changed pixels in a full-page Tailwind CSS capture

Two modes, four outputsChoose pixel or perceptual comparison, then overlay, heatmap, mask or metadata only. Region-awareInserted or removed sections are compared independently, so downstream layout shifts do not become page-wide noise. Direct visualizationOverlay, heatmap and mask assets are returned through the same direct storage contract.

Monitor the page.
Act on the change.

Capture important pages on a schedule, compare each successful run, and send incident events to the systems that need to act. Create the monitor once; Scrinly handles the capture, baseline, diff, and recovery lifecycle.

Rolling or fixedRolling monitors advance after a successful capture and diff. Fixed baselines change only when you replace them. Incident-awareVisual changes and repeated capture failures open and recover independently. Additive creditsThe baseline costs one screenshot, plus regions when enabled. Later runs add one diff credit.

The page’s visual system,
as structured data

Understand a page’s visual language without manually inspecting its CSS. Use structured colors, typography, spacing, motion, and layout signals in design audits, QA, and downstream agent workflows.

Palettefrequency ranked
#5E6AD2
#08090A
#F7F8F8
#8A8F98
#E8FF5A
#26282D
AaInter Variable · 400 500 590
AaSF Mono · 400
Aaself-hosted · woff2 · swap

Deterministic evidenceThe same observed values are sorted, sanitised and capped consistently. No model calldesign:true stays inside the capture flow. Only a requested Style Guide invokes OpenAI. Coverage declaredCaps, truncation and omitted counts stay visible instead of disappearing.

Turn a full page into
model-sized regions

Give vision models a long page in readable pieces instead of one compressed image. Agents can inspect navigation, content, calls to action, and the footer while preserving document order and page context.

A real capture of Tailwind CSS . Every crop, bound, hash and byte count in this panel comes from the saved API response. Scrinly is not affiliated with Tailwind Labs.

One item, many parts Cuts carry evidence Gaps are declared

Add a style guide
only when you need it

Turn a captured page into a reusable visual reference for design reviews, brand audits, and agent handoffs. Each section distinguishes what was observed, what was inferred, and where the available evidence is incomplete.

Visual foundation

observedA dark neutral canvas with a cool blue accent, high-contrast display type and a restrained radius system. The values are tied to the screenshot and evidence-part sha256 values.

Components and states

observedCards share one border treatment. Buttons repeat the same accent, radius and focus treatment.

Motion

observedCards reveal at 600ms with a 100ms stagger.

inferredThe repeated timing curve may represent one shared transition token. The inference is labelled, not promoted to an observed fact.

Responsive behavior

Observations cover the captured breakpoints only.

Evidence gaps

Truncated evidence and omitted values are reported explicitly.

Guaranteesverified per response

Every literal is checked. Each color, duration, easing, breakpoint and token literal is matched back to the screenshot projection. Excess unknown values, a missing heading or truncated output fails the add-on instead of publishing it.

Observed and inferred are labelled. Inference is allowed but clearly separated, so an agent can distinguish measured evidence from model interpretation.

The screenshot remains available. A provider outage returns a partial result, keeps the successful screenshot and refunds only the Visual Style Guide increment.

Provider input stays bounded. The model receives screenshot crops and a sanitised projection—not the target URL, hostname, HTML, headers, cookies, selectors or supplied page text. Visible copy and branding can still appear in the pixels.

Built for
AI agents

Connect Codex, Claude Code, and other MCP clients with scoped OAuth. Agents can capture, compare, wait for durable jobs, and check available credits without receiving your API key or learning the full REST API.

Remote MCP · OAuth

Four visual tools.
Scoped access.

Sign in through Scrinly, approve only screenshot, diff, and usage permissions, then let the agent work. Every billable call still declares its maximum credits before work begins.

  • Claude Code
  • Codex
  • Gemini
  • Cursor
Connect with OAuth
Agent skill

The safe workflow
travels with the agent.

The Scrinly skill teaches compatible agents how to choose the right tool, set the smallest valid credit ceiling, preserve image and region integrity, and poll one accepted job without paying for a duplicate. It contains no credentials and cannot expand the scopes you approved. 

npx skills add davmixcool/skills \
  --skill scrinly \
  -g
Install the agent skill
OAuth by defaultInteractive clients use browser authorization and scoped consent. No Scrinly API key is copied into the client. Credits stay explicitTool discovery, usage checks, and status polling are free. Billable tools require maxCredits. OAuth or API keyClients without OAuth and backend services can use a bearer API key; images remain direct provider resource links.

Know exactly what
you are spending

Credits are charged before the work starts and refunded when it fails, so a request you never got is a request you never paid for. Every response carries your balance in the headers.

Free to start100 credits every month, no card. Paid plans from $29. Yearly saves 20%Billed once, allowance still refreshes every month. Refunded on failureA failed capture costs nothing. Partial failures refund the add-on that failed.

Simple credit pricing

A screenshot costs one credit. regions adds two and a platform styleGuide generation adds three; a cached Style Guide or BYOK adds one. A visual diff is its own request at one credit. Plans differ only by monthly allowance and operating limits.

Switch to yearly and save up to $1,200 a year.

Free

Everything, at a small scale.

$0/mo

Free forever

Start free
  • 100 credits a month
  • No recurring monitors
  • Stored screenshots
  • Regions & Visual Style Guide
  • Monthly allowance can run up to:
  • 100 screenshots
  • 33 captures with regions
  • 16 fresh platform Visual Style Guides
  • 100 visual diffs

Startup

For a first production workload.

$29/mo

billed monthly

Choose Startup
  • 2,000 credits a month
  • 30s render timeout
  • 10 monitors · daily minimum
  • Screenshot batches & webhooks
  • Monthly allowance can run up to:
  • 2,000 screenshots
  • 666 captures with regions
  • 333 fresh platform Visual Style Guides
  • 2,000 visual diffs
Most popular

Pro

The plan most teams settle on.

$99/mo

billed monthly

Choose Pro
  • 10,000 credits a month
  • 60s render timeout
  • 50 monitors · hourly minimum
  • Priority capture queue
  • Monthly allowance can run up to:
  • 10,000 screenshots
  • 3,333 captures with regions
  • 1,666 fresh platform Visual Style Guides
  • 10,000 visual diffs

Business

Steady, high-volume pipelines.

$199/mo

billed monthly

Choose Business
  • 30,000 credits a month
  • 60s render timeout
  • 200 monitors · 15-minute minimum
  • Ephemeral provider BYOK
  • Monthly allowance can run up to:
  • 30,000 screenshots
  • 10,000 captures with regions
  • 5,000 fresh platform Visual Style Guides
  • 30,000 visual diffs

Prime

Maximum capacity for high-volume screenshot and monitoring workflows. Everything in Business, with three times the allowance and storage on your own bucket.

  • 90,000 credits a month
  • 60s render timeout
  • 1,000 monitors · 5-minute minimum
  • Custom storage buckets
  • Monthly allowance can run up to:
  • 90,000 screenshots
  • 30,000 captures with regions
  • 15,000 fresh platform Visual Style Guides
  • 90,000 visual diffs
$499/mo

billed monthly

Choose Prime

Capacity figures are separate ways to spend one month’s full allowance, rounded down. Cached Visual Style Guides and BYOK can cost less.

Yearly plans are billed once but the credit allowance still refreshes every month — an annual plan is not twelve months of credits handed over at once. Credits are charged before a render starts and refunded if it fails. See how credits work →

The questions that
come up first

Everything here is how the API behaves today, not a roadmap. Shipping history lives in the changelog.

Do I have to run any browsers?

No. Renders run on the cloud, so there is no Chromium to operate, no container to size and no queue to babysit. You send a URL and get bytes or JSON back.

What pages can I capture?

Public pages support fullPage, theme and format controls. You are responsible for having the rights and permission required to capture the target.

Safe visual options such as viewport, device, theme, format, waiting, full-page capture and explicit cleanup work on public pages.

cookies, authorization headers and browser actions are available for pages you are authorized to access; the obligation is yours, not a gate we can check for you. Private network targets, loopback, link-local and reserved destinations are blocked outright, and overriding robots.txt requires a domain you have verified — see automation transparency.

Synchronous or queued?

Synchronous by default: one request, one response. Pass async=true or a webhookUrl and the job is queued instead — you get a 202 with an id to poll, or the result delivered to your endpoint.

A platform-funded Style Guide can be queued with its screenshot. BYOK stays synchronous, because a provider credential is never persisted or queued. Screenshot batches are screenshot-only — regions and styleGuide are rejected there, since a batch reserves exactly one credit per item.

Where do the images go?

Stored screenshots, region crops and diff visualizations return direct provider URLs from Scrinly B2 or your B2, S3 or R2 bucket. Assets used for diffing and monitoring must be publicly fetchable; retention follows the selected storage destination.

What happens when something fails halfway?

You keep the successful screenshot and pay only for the completed work. A failed regions or Visual Style Guide add-on returns a partial result and refunds that add-on.

If the browser capture itself fails, the complete charge is refunded. Component errors stay typed so an agent can decide whether to retry.

Can I use my own AI provider key?

Yes, in X-Scrinly-Provider-API-Key. It makes Style Guide generation one credit and the key is never stored, echoed or logged. Textual provider input excludes the URL, hostname, HTML, supplied page text and credentials. The model still sees visible copy and branding inside the screenshot crops.

One credit for the screenshot. Add-ons are charged once, and only the component that fails is refunded.
Credit metering One atomic statement, safe under concurrency
API jobs, monitor records and status routes are account-scoped. Direct storage URLs follow the access policy of their B2, S3 or R2 destination.
Tenant isolation Owned API records return 404 across accounts

Give your agents
visual context they can trust.

Start free Read the docs

100 credits a month. No card, no sales call.