Annotate screenshots before pasting them into any AI chat.
Copy a screenshot, press ⌘⇧A in VS Code, mark it up, and paste it straight into your agent. Snapmark lives on your clipboard, so it works with every AI chat that takes an image.
Open source · MIT · Nothing leaves your machine
Paste the result into
…and anything else that accepts pasted images.
The whole workflow
How Snapmark works
Three moves, and no new app to learn. Snapmark never touches your agent’s window: it works through the clipboard, so it fits every tool you use.
- 1
Copy a screenshot
Use the screenshot tool you already have: ⌃⇧⌘4 on macOS, Win+Shift+S on Windows, any snipper on Linux. A pencil lights up in VS Code’s status bar.
Screenshot copied to clipboard⌃⇧⌘4 - 2
Annotate in VS Code
Press ⌘⇧A (Ctrl+Shift+A). The editor opens on your screenshot: hide what’s private, number the steps, point, comment, measure.
main✎ Annotate12Copy⌘⇧A - 3
Paste into any AI chat
Hit Copy, then paste into Claude Code, Copilot Chat, Cursor, Codex — anything that takes a pasted image. Big Retina shots are shrunk on the way so they don’t eat tokens.
Why does this reset after Save?
Agent · Looking at step 2: the Save handler re-reads the form before the toggle’s state is committed…Message your agent…⌘V
Made for AI chats
Three things the marketplace doesn’t ship.
Plenty of extensions edit images. Snapmark is built for one job: getting a screenshot into a model’s head exactly the way you see it.
Blur sensitive regions
Hide API keys, tokens and personal details before the image leaves your machine. Snapmark covers them with a solid block, so there’s nothing to un-blur.
Numbered step callouts
Drop 1 · 2 · 3 markers to walk the model through a UI flow in order. No more “what am I looking at?” back-and-forth.
Auto-compress on copy
Retina screenshots are resized to 1920px on the long edge when you copy, so vision models don’t spend tokens on pixels they can’t use.
The toolkit
Everything you reach for, one key away.
Go on, play with them. These are the real tools, just smaller.
Compare
Put the design next to what shipped, labelled Actual and Expected, for “make it look like this”.
Spotlight
Dim everything except the part that matters.
Measure
Drag the box. Distances in real CSS pixels.
Color picker
Label any pixel’s exact hex.
Comment bubbles
Figma-style notes that point at a spot.
Arrows
One-way or two-way, heads that scale.
Highlight
A translucent marker over anything.
Shapes & pen
Boxes, ellipses and freehand.
Text
Type labels anywhere on the image.
Crop
Trim to just the part you mean.
Paste images in
Drop a second image onto the canvas.
Your toolbar, your way
Dock it on any edge, reorder the tools.
Private by design
Your screenshots never leave your machine.
The image goes from your clipboard, through Snapmark, and back to your clipboard. That’s the whole trip. You decide where it goes after that.
- No network calls
- Snapmark never sends anything anywhere.
- No telemetry
- No analytics, no usage pings, no crash uploads.
- No account
- Install it and go. Nothing to sign in to.
- Open source
- MIT licensed. Read every line on GitHub.
One more thing
And yes, it comes with a pet.
A tiny pixel cat lives in the corner of the editor. It naps, grooms, wanders off now and then, and never ends up in your screenshots. Prefer a dog? Right-click to swap. Prefer neither? One setting.
Go on, click them.
Stop describing your screen. Show it.
Free and open source. Install it, copy a screenshot, press ⌘⇧A.
