How to report a bug to Cursor (and give it a screenshot)

Cursor's agent is good at finding code once it knows what to look for. The hard part is the hand-over. You can see the broken button. Cursor only sees text and files. A good report bridges that in one message, so the agent doesn't have to guess.

Step by step, by hand

1. Reproduce the bug once more. Note exactly what you clicked. "I clicked Save" is more useful than "saving is broken".
2. Screenshot the broken part. Draw a box or circle on it if the picture shows more than one thing.
3. Add it to Cursor's chat. Cursor's docs say you can drag and drop an image file into the chat input, or paste it with Cmd+V (Ctrl+V on Windows).
4. Write what you expected. One sentence, in plain words. "The Subscribe button should be blue like the other buttons."
5. Add the clues. The page address, and red text from the browser console if there is any. Open it with Cmd+Option+J on a Mac or Ctrl+Shift+J on Windows.
6. Say the scope. "Just this button" or "every button like this one". More on that in why your agent asks if a change is local or global.

Put together, it reads like this:

[screenshot with the button circled]
Page: localhost:5173/pricing
The Subscribe button is grey and looks disabled.
Expected: blue, like the Start trial button above it.
Fix it only on this page.
Console: no errors.

Tips that help Cursor most

• Name things the way the page does. If the button says "Save plan", write "Save plan". The agent can search for that exact text.
• One bug per message. Three bugs in one message often get one fixed and two forgotten.
• Compare to something that works. "Should look like the Start trial button" points the agent at code it can copy.
• Copy error text, don't screenshot it. Text is exact and searchable. If you only have a picture of an error, Image to text pulls the words out.
• If it says "fixed" but you still see it, reload the page and check you're on the right address before reporting again.

Cursor's own browser

Cursor also has a built-in browser that you can attach to a prompt with @Browser. That's handy when the agent should open a page and look for itself. For a bug you spot while using your app in your normal browser, a screenshot or report is still the quickest way to show it exactly what you saw.

The one-key way with Darnly

Darnly is a free Chrome extension that writes the report for you. Press Option+Shift+B (Windows is coming soon), click the broken part, type what you expected and press Enter.

If your project is on your computer, link its folder once and click Set up this project. Darnly adds a note to AGENTS.md that tells Cursor's agent where bug reports live, and a /darnly command. Each bug becomes a markdown file and a PNG in .darnly/inbox/: the screenshot with the broken part boxed in red, the element's selector, text, HTML and styles, the page address, and console errors and failed requests from the last 60 seconds. Then in Cursor you type /darnly, and the Darnly panel tells you when each bug is fixed.

If your code lives in the cloud, skip the folder. After you save a bug, click 1. Copy picture, paste it into Cursor, then 2. Copy details and paste that too. Or add the Darnly MCP server to Cursor once, and its agent fetches your bugs itself: set up Darnly in Cursor.

Darnly doesn't make the agent smarter. In our tests, a good screenshot and sentence got bugs fixed just as often. What it saves is the hand-over: no app switch, no hunting for the console, and the exact element and errors kept for you. See Darnly vs a plain screenshot for the numbers.