You can now generate images directly from Cursor’s agent!
How it works
Describe the image in text or upload a reference to guide the underlying image generation model. Images are returned as an inline preview and saved to your project’s assets/ folder by default.
Useful for
Creating UI mockups
Generating product assets
Visualizing architecture diagrams
We’d love your feedback!
What kinds of images are you generating?
How does this fit into your design-to-code workflow?
What image generation improvements would be most helpful?
If you’ve found a bug, please post it in Bug Reports instead, so we can track and address it properly, but also feel free to drop a link to it in this thread for visibility.
Seems a bit random right now whether or not it works, and local network conditions / our backend contributes to that randomness. We’re working on a fix.
I’m having a lot of trouble with the image_generator tool. It frequently times out, which causes the entire chat to freeze, and I’m also currently hitting 429 errors.
Even when it does work, there are persistent issues with saving files. The AI tries to specify the path manually—sometimes with a filename, sometimes without—but it results in very strange behavior.
I only managed to successfully create one test image, but it was saved with a bizarre filename that included the entire absolute path as part of the name: "C:\Users\qwerty\.cursor\projects\c-Users-qwerty-folder-project\assets\diamond_icon.png"
Did anyone actually manage to generate an image? I spent a ton of money on tokens trying to get it to generate some image, nothing works at all. Same as always, Cursor. Thanks for rolling out a broken build. same as always)
On the latest early access version, it works great for me, I have to explicitly ask for an image, but so far has worked every time. What model are you using? I did not find anything on image gen in docs, but for example sub agents require max mode on legacy pricing plans, could this possibly be the case for you?
yes, I’m on legacy pricing plan, but is this really a reason to disable some functions for me, why not spend 0.1 requests for example on image generation or something like that
My opinion is harsh but honest - it’s a bells and whistles feature few asked for.
2.3 was the right direction fixing bugs and improving stability, please focus on fixing hooks, MCP issues and addressing all the agent view feedback and all the other various issues across the forum
I know it’s a Karen opinion but I pay $200 for this
Just to keep you updated, we’ve been in talks with our provider to reduce 429 errors in image generation, and things are moving in the right direction!
Hey @Colin - thanks for keeping us up to date. Contrary to what someone said above - I think it’s a very useful feature - I don’t have to go elsewhere in my dev environment when I need to gen assets!
Just a +1 here that I’m on v 2.3.35 and my agent (auto, agent mode) isn’t aware of this skill/tool (even after I’ve sent it this URL :D)
@Colin Oops
Coincidentally, I got the update to 2.4.28 today. The skill looks like it executed, but I get this
I’m uploading in my normal root dir of a project where I normally work, and I already have an assets/ folder.
