Urlbox

Screenshot a website

render_screenshot

Destructive

Take a screenshot of a URL or of HTML (HTML to image): full page, a single element, retina, dark mode, without ads or cookie banners. Returns the image and its URL. One render can also save the page's markdown, HTML and metadata: set save_markdown, save_html or save_metadata instead of making a second call. The markdown and metadata come back with the result. Any other Urlbox option (proxies, cookies, headers, user agent and many more) goes in options: look it up with describe_options.

Tool call template

Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {
      "intent": "<intent>"
    },
    "name": "render_screenshot"
  }
}

Input Schema

Name	Required	Description
`gpu`	No	Enable GPU acceleration to render 3D scenes and heavy WebGL content. This is a beta feature and requires pre-approval. Please contact support@urlbox.com to enable this feature on your account. Requires the ultra plan or above.
`url`	No	The URL or domain of the website you want to screenshot. We will automatically prepend http:// if it is missing.
`html`	No	HTML to render instead of a URL (HTML to PDF, HTML to image).
`delay`	No	The amount of time to wait before Urlbox captures a render in milliseconds.
`force`	No	Generate a fresh render on each request, instead of getting a cached version.
`width`	No	The viewport width of the browser, in pixels.
`format`	No	Output format: png, jpeg, webp. Defaults to png.
`height`	No	The viewport height of the browser, in pixels.
`intent`	Yes	Required. In one or two sentences, what you are trying to achieve for the user with this call and why (for example: 'Screenshot the pricing page so the user can compare plans'). Recorded for analytics and audit; do not include secrets.
`retina`	No	Take a 'retina' or high-definition screenshot, equivalent to setting a device pixel ratio of 2.0 or @2x. Please note that retina screenshots will be double the normal dimensions and will normally take slightly longer to process due to the much bigger image size.
`options`	No	Any other Urlbox render option by name, e.g. { "gpu": true }. For an option this tool doesn't list, look it up with describe_options first (it covers every option); don't search the web. Validated by the Urlbox API.
`project`	No	Project id (proj_…) or exact name to render with. Defaults to your default project.
`quality`	No	The image quality of the resulting screenshot (JPEG/WebP only) Range: 0-100.
`selector`	No	Take a screenshot of the element that matches this selector. By default, if the selector is not found, Urlbox will take a normal viewport screenshot. If you prefer Urlbox to fail the request when the selector is not found, pass fail_if_selector_missing=true.
`wait_for`	No	Waits for the element specified by this selector to be present in the DOM before taking a screenshot or PDF. By default, Urlbox will take a screenshot or PDF if the wait_for element is not found after waiting for the time specified by the wait_timeout option. If you prefer Urlbox to fail the request when the wait_for element is not found, pass fail_if_selector_missing=true
`block_ads`	No	Blocks requests from popular advertising networks from loading.
`dark_mode`	No	Emulate dark mode on websites by setting prefers-color-scheme: dark
`full_page`	No	Specify whether to capture the full scrollable area of the website. For PDFs, full_page mode will attempt to capture the whole website onto one single page PDF document. It's likely you'll want to also hide any cookie banners that crop up during a full page screenshot, so we recommend you use click_accept and hide_cookie_banners too.
`save_html`	No	Also save the page's rendered HTML.
`wait_until`	No	Waits until the specified DOM event has fired before capturing a render. The available options are: domloaded (the DOMContentLoaded event is fired); mostrequestsfinished (consider navigation to be finished when there are no more than 2 network connections for at least 500 ms); requestsfinished (there are no more than 0 network connections for at least 500 ms); loaded (the load event is fired); One of: domloaded, mostrequestsfinished, requestsfinished, loaded.
`thumb_width`	No	The width of the generated thumbnail, in pixels. Omit for a full-size screenshot.
`click_accept`	No	Similar to the hide_cookie_banners option, but instead of hiding the banners, this option attempts to click on the 'Accept' button, in order to accept cookies.
`thumb_height`	No	The height of the generated thumbnail, in pixels. Omit for a full-size screenshot.
`wait_seconds`	No	How long to wait for the render before returning. 0 queues it and returns a render id at once. If the wait runs out the render keeps going; fetch it with get_render.
`hide_selector`	No	Hide specific HTML elements on the page before rendering. This option accepts a comma-delimited string of CSS element selectors that will be hidden by setting their style to visibility: hidden !important; pointer-events: none !important;. This preserves the page layout while making elements invisible. This is particularly useful for: Hiding pop-ups, banners, or cookie notices; Removing advertisements or promotional overlays; Excluding navigation menus or sidebars; Hiding specific content sections; Selector types supported: Element selectors (e.g., h1, div, img) - Hide all elements of that type; Class selectors (e.g., .popup, .banner) - Hide elements with specific CSS classes; ID selectors (e.g., #header, #sidebar) - Hide elements with specific IDs; Complex selectors (e.g., .nav ul li, div.content > p) - Use any valid CSS selector; Multiple selectors - Combine multiple selectors with commas; Tip: To find the selector for any element, open your browser's DevTools, right-click the element in the Elements tab, and select "Copy > Copy selector".
`save_markdown`	No	Also extract and save the page as Markdown alongside the render.
`save_metadata`	No	Also extract and save page metadata (title, description, OpenGraph, etc.).
`engine_version`	No	Sets the version of the urlbox rendering engine to use when rendering the page. This can be useful for testing how a page will render in the latest version of our rendering engine. One of: stable, latest, experimental.
`hide_cookie_banners`	No	Automatically hides cookie banners from most websites, by setting their style to display: none !important;

Schema Changelog

Changes observed during successful MCP inspections.

AddedNov 9, 2026

TDQS

A4.1/5.0

Behavior3/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations already declare openWorldHint=true, destructiveHint=true and idempotentHint=false, so the safety profile is structured. The description adds genuine behavior beyond that — the return shape ('the image and its URL') and that markdown/metadata come back in the same result — but says nothing about the fresh-vs-cached/repeat-call implications that the non-idempotent and destructive hints raise.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Front-loaded with purpose and capabilities, followed by return behavior and the options escape hatch — a sensible ordering. The sentence 'The markdown and metadata come back with the result' is slightly redundant with the preceding clause, which keeps it from a 5.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a 29-parameter tool with a nested `options` object and no output schema, the description covers what an agent genuinely needs beyond the schema: what is returned, the combined-render shortcut, and where to find other options. Missing only guidance on async behavior (wait_seconds=0 → get_render) and repeat-call semantics.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 100%, so the baseline is 3, but the description adds real meaning on top: it explains the purpose of the save_markdown/save_html/save_metadata trio relative to separate render calls, and clarifies that `options` is the escape hatch for the many unlisted Urlbox parameters. It does not touch the large set of rendering parameters (selector, wait_for, full_page), leaving those to the schema.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

States a specific verb and resource ('Take a screenshot of a URL or of HTML'), then enumerates the distinctive capabilities (full page, single element, retina, dark mode, ad/banner blocking). It also distinguishes itself from sibling calls by noting that save_markdown/save_html/save_metadata avoid 'making a second call', which implicitly routes the agent away from render_markdown/render_html.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Gives concrete routing advice: use the save_* flags instead of a second call, and send unlisted Urlbox options through `options` after looking them up with describe_options. It stops short of explicitly contrasting itself with render, render_html, or render_pdf when the goal is only markdown/HTML/PDF output.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

All Urlbox tools Inspect

Try in Browser

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Glama MCP Gateway

Add one secure layer between your agents and this server.