get_snapshot

Let the agent see a board, a section or a Frankenstein page without a browser - small thumbnails with positions, or a page's parts in row order with their crop rects. A parts view, not a rendered image.

Plan	Quota	Changes data	Hints
All	None	No	read only

Inputs

Name	Type	Required	Limits	What it does
`canvasId`	string	Yes	1 to 100 characters	The board
`sectionId`	string	No	a section id from `get_canvas`	Show this section’s references with their positions
`stitchId`	string	No	a Frankenstein page id from `get_canvas`	Show this page’s parts, row by row
`thumbs`	boolean	No	default `true`	`false` returns the text lines only

Pass a sectionId or a stitchId, not both. With neither, you get the board’s sections, Frankenstein pages and first references.

A Frankenstein page

{ "name": "get_snapshot", "arguments": { "canvasId": "c3f1a0d2-…", "stitchId": "st_…" } }

The result is text lines with an image after each one:

Frankenstein page "Checkout A" (st_…, mobile, collage), 2 rows top to bottom.
This is a parts view, not a rendered composite: each part below is its full source image, and its crop rect says which band of it the page uses.
Row 1: Nav (part pt-…) from dribbble.com https://dribbble.com/shots/… Crop x 0.00, y 0.00, w 1.00, h 0.12 (fractions of the image below; the page uses only this band)
[image]
Row 2: Hero (part pt-…) from behance.net https://www.behance.net/gallery/… Crop x 0.00, y 0.10, w 1.00, h 0.35 …. Note: steal: the big date
[image]
Build each row from what you see here; check your build against these images.

Builds placed next to the page with add_to_canvas (type build) are listed at the end with their links.

How the thumbnails work

The server fetches each image at its source’s own small size (Pinterest 236x, Dribbble resize=400x300, Behance disp) and hands the bytes to the agent in the result. They are never stored, cached, logged or served at any address.
Only image hosts on the search allowlist and Mobbin’s image hosts are fetched. Your uploads and build screenshots are named with their link but not fetched.
Up to 12 images per call, about 3 seconds each, about 1.5 MB in total. An image over 300 KB with no small size is skipped, and the line says so.

Messages

Message	Fix
“That Frankenstein page was not found…”	Read the page ids with `get_canvas`
“That section was not found…”	Read the section ids with `get_canvas`
“Pass a sectionId or a stitchId, not both.”	Pick one