Describe the Bug
I have observed that some of my successful AI requests are being recorded with zero token usage. I understand that cancelled or failed requests would correctly show zero tokens, but the instances I’m referring to are separate from these. The interactions completed successfully in the chat interface, providing a response without any visible errors.
As an example, I have attached a screenshot of my usage history which highlights several successful requests logged with zero tokens, alongside a few that errored (which are expected to be zero).
Could you please clarify if this is intended behaviour? For instance, I was wondering if certain models used in ‘Auto’ mode are not metered, or if this might be a bug in the token calculation or reporting system?
While this issue doesn’t negatively impact my user experience, I wanted to bring it to your attention for the sake of data accuracy and in case it points to a different underlying issue.
Thank you for your time and for the fantastic work you do on Cursor. Please let me know if I can provide any further information.
Steps to Reproduce
I do not do anything special to reproduce it. It happens during normal chat interactions in Agent mode.
Expected Behavior
All successful AI requests should have their token usage accurately calculated and reflected in the usage statistics. This ensures the reliability of the entire metering system and builds confidence that paid requests, such as those using personal API keys or frontier models, are also being tracked with complete accuracy.
Screenshots / Screen Recordings
Operating System
Windows 10/11
Current Cursor Version (Menu → About Cursor → Copy)
Version: 1.2.4 (user setup)
VSCode Version: 1.99.3
Electron: 34.5.1
Chromium: 132.0.6834.210
Node.js: 20.19.0
V8: 13.2.152.41-electron.0
OS: Windows_NT x64 10.0.26100
Does this stop you from using Cursor
No - Cursor works, but with this issue