Image pasting support on Linux in Cursor CLI

Describe the request

I cannot past images into the Cursor CLI. GPT5.2 summary of what’s going on with the clipboard on my computer:

Arch Linux (X11), CopyQ 13.0.0. When an image is in the clipboard, xclip -selection clipboard -t TARGETS -o includes image/png (and others, plus application/x-qt-image):

$ xclip -selection clipboard -t TARGETS -o
application/x-qt-image
image/png
image/bmp
image/cur
image/icns
image/ico
image/jfif
image/jp2
image/jpeg
image/jpg
image/pbm
BITMAP
image/pgm
image/ppm
PIXMAP
image/tif
image/tiff
image/wbmp
image/webp
image/xbm
image/xpm
TARGETS
MULTIPLE
TIMESTAMP
SAVE_TARGETS

Also confirmed the clipboard data is actually readable as valid PNG bytes:

$ xclip -selection clipboard -t image/png -o >/tmp/clip.png && file /tmp/clip.png
/tmp/clip.png: PNG image data, 963 x 260, 8-bit/color RGB, non-interlaced

So on Linux/X11 the CLI should be able to “paste image from clipboard” by retrieving image/png from the clipboard selection (optionally falling back to image/jpeg/image/webp if image/png is not available).

From conversation here https://x.com/n2parko/status/2009816451428102335

1 Like

Hey, thanks for the detailed feature request and the technical explanation!

This is a known limitation: pasting images in the CLI on Linux isn’t supported right now. The team is aware of it, and it’s on the roadmap for future improvements.

Thanks for the technical context about xclip and clipboard targets too. That info will be helpful when we implement this feature.

1 Like

MCP servers can send images to agents no-problem - plus agents can probably read image files - so one or both of these might be workarounds?

For anyone (like me) who came here looking for a macOS solution, here is a quick workaround: use Shift + Cmd + 4 to take a screenshot. When the preview thumbnail pops up in the bottom right corner, simply drag and drop it directly into the Cursor input. It will paste the file path, and Cursor will be able to read and use the image perfectly.

1 Like

That works yeah!

A trick on top of this is to create a skill that saves the picture to a .screenshots folder within the workspace. You can create a bash, etc. script to create the folder if it does not exist and copy the screenshot from the temporary location.

And as it is in a script with very limited scope it can even be added to the allowlist of the Cursor CLI configuration so that permission is not required every time and no higher privileges are allowed (for mkdir and cp).