I updated to the latest Cursor this evening, and it seems to have lost the ability to take screenshots of the browser content. Earlier today it was doing things like this:
Thought for 2s
Performed 2 browser actions
Navigated to http://localhost:3000/…
Thought for 3s
Captured screenshot
Thought for 7s
The content is displaying correctly.
But now after the update it says:
Regarding screenshots - looking at my available tools, I don’t have a browser_take_screenshot tool. I only have browser_snapshot which captures the accessibility tree (semantic structure), not a visual screenshot. The screenshot capability either isn’t available in my current toolset or was removed. The accessibility snapshot is useful for verifying interactive elements but misses content that doesn’t have proper semantic roles.
The manual screenshot tool exists in the browser address-bar “…” menu, and works correctly when I use it there, and I can paste that image into chat for the agent to examine, but it can no longer take screenshots itself.
Steps to Reproduce
Ask the an IDE agent to check something visual about a webpage using the IDE browser tools.
Expected Behavior
IDE chat agent should be able to use the browser-screenshot tool to examine browser rendering details.
Operating System
MacOS
Current Cursor Version (Menu → About Cursor → Copy)
I got the exact same problem - on 2.2.14 - however it worked yesterday, I think I have updated in the mean time (so the problem was possibly introduced with 2.2.14 at my end).. I am on windows
I tried OPUS 4.5 (max) and composer 1 (max) with a restart of cursor in between
I have updated to 2.2.14, and still experience the error. This is what the agent says:
Looking at my available browser tools, I don’t see a visual screenshot capability - I only have browser_snapshot which returns the accessibility tree (the semantic snapshot you see above).
The tools I have for the browser are:
browser_navigate - navigate to URL
browser_snapshot - capture accessibility tree
browser_click/type/hover/select_option/press_key - interact with elements
It appears the visual screenshot tool (browser_take_screenshot) that was available yesterday is not currently in my toolset.
(incidentally, the “which you see above” comment from Claude is apparently referring to an embedded response from the browser_snapshot tool, which Claude can see as part of the chat, but is hidden from me by the UI – not a bug, per-se, just interesting to observe how Claude’s view of the chat differs from mine…)
I’ve updated to 2.2.17, and the issue persists. Here is what Claude has to say:
The browser navigation works and I get the accessibility snapshot, but I still don’t see a visual screenshot tool in my available toolset. My current browser tools are:
There’s no browser_take_screenshot or similar. Either:
The screenshot tool hasn’t been re-enabled yet in this update
It might require a different configuration or feature flag
It may have been replaced with a different mechanism
The semantic snapshot works fine though - I can see the page structure and interact with it. Is there anything specific you’d like me to verify visually that you could check manually, or shall we continue with other work?
Can you see how much space is visible on the left side of the “Lightweight Spaceship” text versus the right side inside the selection-element? Is it visually-balanced?
I can’t visually verify the spacing without the screenshot tool - the accessibility snapshot only gives me semantic structure, not visual layout information.