Screenshot API

Screenshots of the page. Not the cookie banner.

A screenshot API that removes consent walls, sign-in modals and paywall overlays before it captures. Most screenshot APIs succeed 93% of the time and return something usable 69% of the time — we measured it, and fixed it.

Try it on any URL Read the API

93%renders succeedacross a 29-URL benchmark

69%are actually usablethe gap nobody quotes

1 / 29blocked by bot protection7 ruined by consent walls

~2smedian renderp50, viewport capture

See the difference

Left is the page as it loads. Right is what the API returns. No signup, no key.

overlay_mode=off raw

overlay_mode=dismiss clean

The gap nobody quotes

We benchmarked 29 URLs across headless, stealth and headful browsers, from both a residential line and a datacenter address. Every number on this page came out of that run.

Clean by default

Consent walls, sign-in modals, newsletter popups and paywall scrims are removed before capture. Known CMPs are declined properly; anything else is caught by shape — fixed, painted, high z-index, covering the page.

It never clicks Accept

Declining on your behalf for a throwaway render is defensible. Agreeing on your behalf is not. Reject buttons that are really paywall funnels are detected by their own label and skipped in favour of hiding.

No signup to try it

The examples on this page work pasted, with no key and no account — 20 renders an hour against your address. A free key raises that, and signed URLs let you drop a render into an <img src> without publishing a credential to every browser that loads the page.

The controls you expect

Viewport, device pixel ratio, full-page, PNG / JPEG / WebP, quality, dark mode, locale, timezone, element waits and delays. Sensible defaults so the simple call stays one line.

Built for agents

An MCP server, an OpenAPI description, llms.txt and a machine-readable api-catalog. Your agent can discover and call it without you writing a wrapper first.

Honest about limits

Some pages are behind bot protection and no browser trick gets past them — we tested that too, and say so rather than pretending. You get a clear error, not a picture of a challenge page.

One call

# No key needed. Paste this as it is.
curl -X POST https://shots.rendershed.com/v1/render \
  -H "Content-Type: application/json" \
  -d '{"url": "theguardian.com/international"}' \
  -o shot.png

# A free key raises the limit from 20 renders an hour to 100 a month
# and keeps working from any address:
#   -H "X-API-Key: $RENDERSHED_KEY"

const res = await fetch("https://shots.rendershed.com/v1/render", {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    // Optional. Without it you are on the anonymous tier.
    ...(process.env.RENDERSHED_KEY && {
      "X-API-Key": process.env.RENDERSHED_KEY,
    }),
  },
  body: JSON.stringify({
    url: "https://www.booking.com",
    width: 1440,
    response: "json",   // base64 instead of raw bytes
  }),
});

const png = Buffer.from(await res.arrayBuffer());

import os, requests

# The header is optional; leave it out to use the anonymous tier.
res = requests.post(
    "https://shots.rendershed.com/v1/render",
    headers={"X-API-Key": os.environ.get("RENDERSHED_KEY", "")},
    json={"url": "https://www.bbc.com/news", "format": "jpeg", "quality": 80},
    timeout=90,
)
open("shot.jpg", "wb").write(res.content)

<!-- signed server-side; no key reaches the browser -->
<img src="https://shots.rendershed.com/v1/render?url=https%3A%2F%2Fstripe.com&width=1200&sig=..."
     width="1200" height="750" alt="stripe.com" loading="lazy">

Parameters

Name	Type	Default	Description
`url`	string	required	The page to render. http and https only.
`width`	int	1440	Viewport width, 200–3840.
`height`	int	900	Viewport height, 200–4320.
`dpr`	float	1	Device pixel ratio, up to 3.
`full_page`	bool	false	Capture the whole document, not just the viewport.
`format`	enum	png	png, jpeg or webp.
`quality`	int	—	1–100, for jpeg and webp.
`overlay_mode`	enum	dismiss	off, hide or dismiss.
`block_ads`	bool	true	Drop known ad and tracker requests.
`dark_mode`	bool	false	Render with prefers-color-scheme: dark.
`wait_for_selector`	string	—	Wait for a CSS selector before capture.
`delay_ms`	int	0	Extra settle time, up to 10000.
`locale`	string	en-US	Browser locale.
`timezone`	string	UTC	IANA timezone.

Built for agents, not just for people

An agent that can see a page is more useful than one that cannot. Everything needed to discover and call this without a wrapper:

/openapi.json OpenAPI 3.1 description
/llms.txt plain-language summary for models
/index.md this page as markdown
/.well-known/api-catalog RFC 9727 catalog
/.well-known/mcp/server-card.json MCP server card

Pricing

Cached renders are free and never count against your quota. They cost us a disk read, so charging for them twice would be charging for nothing.

Free

$0forever

100 / month

Viewport capture
PNG, JPEG, WebP
Overlay removal
Signed URLs
24-hour cache

Starter

$9.99/ month

3,000 / month

Everything in Free
base64 / JSON response
7-day cache
Retina (2x) output
Email support

Growth

$39.99/ month

20,000 / month

Everything in Starter
30-day cache
Higher concurrency
Priority queue
MCP server access

Scale

$99/ month

75,000 / month

Everything in Growth
Full-page capture
4K and 3x output
Overage at $0.001 / render
Self-hosted option

Questions

What makes this different from other screenshot APIs?

Most of them return the page with the cookie banner still on it. We benchmarked a 29-URL corpus: 93% of renders technically succeeded, but only 69% were images you could actually use. Seven were ruined by consent overlays and exactly one by bot protection. rendershed treats that gap as the product.

Does it click “Accept all” on cookie banners?

Never. It either declines, where a genuine reject control exists, or it hides the banner cosmetically, which consents to nothing. Reject buttons that actually lead to a paid subscription are detected and skipped.

Can it screenshot pages behind bot protection?

Sometimes, and we do not claim otherwise. We measured headless, stealth and headful browsers from both residential and datacenter addresses: the browser makes no difference, the IP makes some, and a small number of sites are simply not renderable. Those return an error rather than an image of a challenge page.

How do I use it in an <img> tag?

Mint a signed URL with POST /v1/sign and use it as the src. The signature covers every parameter, so the URL cannot be edited into a different render, and your API key never reaches the browser.

Is there an MCP server?

Yes. Rendershed exposes an MCP server so an AI agent can render a page and look at it as part of a task, plus an OpenAPI description and llms.txt for discovery.

Do I need an API key to start?

No. Calls without a key run on an anonymous tier of 20 renders an hour, metered against your address, so the examples on this page work when pasted. A free key raises that to 100 a month from anywhere; paid plans run $9.99 for 3,000, $39.99 for 20,000 and $99 for 75,000. Cached renders are free at every tier and never count against a quota.