take a screenshot of this page and save it
Three kinds of capture
| What you get | Cost | |
|---|---|---|
| Viewport (default) | What is on screen right now | A single grab, well under a second |
| Full page | The entire scroll view, stitched from viewport tiles | Roughly a second per screenful |
| One element | Just that element's box | A single grab |
Full-page capture has to scroll the page in viewport-sized steps and wait out the browser's two-captures-per-second limit between each. Ask for it when you need what is below the fold, not by default.
Captures do not touch your disk unless you ask
This is the part worth knowing. The image is handed straight back to whoever called for it, so the screenshots an agent takes to see the page for itself leave nothing behind.
A picture you want to keep is a different request. Ask for one and the capture is written to
~/browsentic/screenshot/ at mode 0600, and the result reports the path.
Change where with screenshotDir in config.
Format and size
Defaults are JPEG at quality 80, downscaled so the longest side is at most 1600 px — far smaller and quicker than PNG, which matters because these go to a model. PNG is available when you need lossless or transparency.
Very tall pages
Full-page capture is capped at 48 tiles and a 16 384 px canvas side. Beyond that the bottom is
cut off and the result says truncated: true, rather than silently returning a partial image and
letting you believe it is the whole thing.
See also
- reference/tools.md § page_screenshot — every parameter
- Theming — if the goal is "this page is unreadable", measuring beats capturing
- internals/subsystems.md — how tiles are stitched and who writes the file