High-performance browser automation bridge and multi-instance orchestrator with advanced stealth injection and real-time dashboard.
Screenshot
Capture the current page as an image. The API defaults to JPEG. The CLI
uses PNG when -o ends in .png and JPEG otherwise; an explicit --format
always takes precedence.
# Get raw PNG bytes
curl "http://localhost:9867/screenshot?format=png&raw=true" > page.png
# Capture a specific element (selector supports ref/CSS/XPath/text)
curl "http://localhost:9867/screenshot?selector=%23checkout-button&raw=true" > button.jpg
# Half-size output (quarter the pixels)
curl "http://localhost:9867/screenshot?scale=0.5&raw=true" > page-half.jpg
# Capture the entire scrollable document, not just the visible viewport
curl "http://localhost:9867/screenshot?beyondViewport=true&raw=true" > fullpage.jpg
# Get JSON with base64 JPEG (default)
curl "http://localhost:9867/screenshot"
# Save to server state directory
curl "http://localhost:9867/screenshot?output=file"
Response (JSON)
Default: {"format":"jpeg","base64":"..."}. With annotate=true the body also
carries annotations: [{ref, role, name, tag, box:{x,y,w,h}}]. With
output=file:
{
"path": "/path/to/state/screenshots/screenshot-20260308-120001.jpg",
"size": 34567,
"format": "jpeg",
"timestamp": "20260308-120001"
}
A selector that matches nothing returns the same selector-failure 404 as
other read endpoints.
Useful flags
API Query Parameters
format:jpeg(default) orpng.quality: JPEG quality0-100(default:80). Ignored for PNG.selector: Unified selector to capture one element (e.g.e5,#id,xpath://...,text:Submit).scale: Rescale the output bitmap. Default1, clamped to0.05–4.0.5halves each axis (quarter the pixels). Invalidquality/scalevalues fall back to the defaults.beyondViewport:trueto capture the full scrollable document instead of just the visible viewport. Ignored whenselectoris set. Withannotate=truethe returned box coordinates are document-relative.annotate:trueto overlay numbered ref boxes on interactive elements (or onselectormatches, which then narrows instead of clips) and return anannotationslist.noAnimations:trueto disable CSS animations before capturing.raw:trueto return image bytes directly instead of JSON.output:fileto save to state directory.tabId: Target a specific tab.
CLI
Without -o the CLI saves to screenshot-<timestamp>.jpg (or .png) in the current directory.
-o <path>: Save to a specific path. A.pngextension selects PNG when--formatis omitted; other extensions use JPEG.--format <jpeg|png>: Explicitly select the image format, overriding extension inference.-q <0-100>: Set JPEG quality.-s <selector>: Capture a specific element.--scale <f>: Bitmap rescale (e.g.0.5).--beyond-viewport: Capture the full scrollable document. Ignored when--selectoris set.--annotate: Overlay numbered ref boxes and print a[n] ref role "name"legend to stdout.--tab <id>: Target a specific tab.
When to use pinchtab capture instead
screenshot returns image bytes only. When the model needs to act on refs
in the same turn it reads pixels, use capture — paired
image + accessibility snapshot from the same DOM epoch.