Sometimes an agent doesn’t want a page’s text — it wants to see it: capture the rendered result, archive a state, verify a layout. That means a real browser. Here’s how an agent renders any URL to an image without running browser infrastructure itself.
POST a URL and the screenshot stall renders it in headless Chromium and returns a PNG as base64 JSON, with the final URL and dimensions. Modern pages build their content with JavaScript, so a render shows what a person sees — not the near-empty raw HTML. Pay-per-call over x402, no key.
# Render any page to a PNG in headless Chromium. ~$0.01 USDC over x402, no API key.
curl -s -X POST https://true402.dev/api/v1/screenshot \
-H 'content-type: application/json' \
-d '{"url": "https://example.com", "fullPage": false, "width": 1280}'
# → { url, finalUrl, width, height, bytes, png_base64: "iVBORw0KGgo…" }
# decode png_base64 to get the image.Set fullPage to capture the whole scroll height (up to a cap) or just the viewport, and choose a width from 320 to 1920. The response carries the exact rendered dimensions and byte size, so an agent can decode the image and know what it got.
Rendering attacker-supplied URLs is the most hostile input a service takes, so it is locked down: the URL is validated, every request the page makes is checked, and the browser’s only way out is a filtering proxy that blocks private and internal addresses. An agent can render untrusted pages without the service being turned into a way to reach internal systems.
An unpaid request returns HTTP 402 with payment requirements; the agent pays ~$0.01 USDC on Base and retries — no signup, no infrastructure. Want clean text instead of an image? Use web-extract. New to the protocol? What is x402?
It calls a rendering endpoint that runs headless Chromium for it and returns the image. true402’s screenshot stall does this: POST a URL to /api/v1/screenshot and it renders the page in a sandboxed browser and returns a PNG as base64 JSON, with the final URL and image dimensions. There is no browser to install, no infrastructure to run, and no API key — the agent pays per call over x402.
Yes. The screenshot stall is a keyless, pay-per-call render endpoint. An unpaid request returns HTTP 402 with signed payment requirements; the agent pays and retries. You can request a full-page capture and set the viewport width (320–1920). Because the page is rendered rather than just fetched, JavaScript-built pages come back the way a user would see them.
Fetching HTML gives you the source, not the rendered result — modern pages build much of their content with JavaScript, so the raw HTML is often nearly empty. Rendering in a real browser produces what a person actually sees: layout, images, and JS-built content. Use true402’s web-extract stall when you want clean text, and the screenshot stall when you need the visual render as an image.
It has to be treated as hostile, and it is. The screenshot stall validates the URL and runs the browser in a sandbox that cannot reach private or internal network addresses — every request the page makes is checked, and the browser’s only path out is a filtering egress proxy that blocks internal ranges. So an agent can render attacker-supplied URLs without the service leaking internal data.