Show your agentwhat you mean.
Record and talk it through. The transcript copies to your clipboard — and your agent can see the frames and files for itself.
Record your screen.
Here's what your agent actually receives:
Transcript · everything pinned in place
00:08walking through the checkout redesign
00:12grab this icon figma.com/…/Checkout-v3 for the cart button
00:19see how it breaks on mobile checkout-mobile.png
00:29keep the desktop grid just like this 0:34
Frames · it can see
Attachments · it can open
A video burns tokens. Text loses the point.
The two obvious options each cost you something. Screen gives an agent the signal — your words, your files, the exact moments — without either tax.
Send a raw videoCostly & vague
- ×The agent decodes every frame — tokens spent before it even starts.
- ×No way to pin a note or a link to an exact second.
- ×Your files and references live somewhere else entirely.
Type it all outSlow & lossy
- ×Slow to write — and you end up explaining it three times.
- ×Words can't hold a layout; design intent dies in translation.
- ×The longer it gets, the more precision you lose.
ScreenPrecise & cheap
- ✓A transcript reads for a fraction of what a video costs.
- ✓Notes and links pinned to the moment you said them.
- ✓Frames only where they matter — the agent still sees.