Aborted/Errored Requests Are Being Counted

Describe the Bug

When using the AI features in Cursor AI Code Editor, even if the AI response is aborted by the user (marked as “Aborted, Not Charged”) or the AI encounters an error due to network issues (marked as “Errored, Not Charged”), these operations still lead to the consumption of the user’s “Included Requests” quota and are recorded in the “Usage Events” list.

This behavior differs from previous Cursor versions, where only successful AI interactions counted towards the “Included Requests” quota. This results in the user’s quota being wasted on non-productive operations that yield no actual output.

Screenshot Explanation:
Due to posting limitations (allowing only one image attachment), all relevant screenshots have been consolidated into a single attached image. This image has only been modified by blurring personal sensitive information (e.g., email ID) and adding red lines to highlight key areas; no other alterations have made. Should there be any doubts or a need for further verification, I am willing to provide additional assistance or further authentication.

  • Images 2, 4, and 6 continuously show an increasing “Included Requests” count.
  • Images 1, 3, and 5 display the detailed “Filtered Usage Events” list. Specifically, images 3 and 5 contain clearly highlighted entries showing “Aborted, Not Charged” or “Errored, Not Charged.” These non-valid operations, however, correspond to an increase in the “Included Requests” count.
  • Furthermore, based on my actual usage observation, “Errored, Not Charged” entries, such as those from “Jun 19, 07:55 PM You Errored, Not Charged” and “Jun 19, 07:54 PM You Errored, Not Charged,” were also counted at the time, although I cannot provide screenshots for these specific records.

Testing Date: This issue was first identified on June 19, 2025, and was re-confirmed to persist on June 20, 2025.

Steps to Reproduce

  1. Open the Cursor AI Code Editor.
  2. To trigger an Aborted operation:
    • Initiate an AI code generation request (e.g., by asking a question in AI chat).
    • Immediately manually stop the AI’s response.
    • Observe if the “Included Requests” count increases.
  3. To trigger an Errored operation (due to network issues):
    • Attempt to initiate an AI code generation request under unstable or interrupted network conditions.
    • Observe if the AI returns an error (displays “Errored”) due to network problems.
    • Observe if the “Included Requests” count increases.
  4. Navigate to the Cursor Dashboard > “Usage” page.
  5. Verification: Check the “Included Requests” count on the “Overview” page and compare it with the “Filtered Usage Events” list.
    • Expected behavior: AI requests that are aborted or error out due to network reasons should not increase the “Included Requests” count, nor should they appear in the usage event list in a way that suggests quota consumption.
    • Actual behavior: Even if marked as “Aborted, Not Charged” or “Errored, Not Charged,” these operations are recorded, and the “Included Requests” count increases.

Screenshots / Screen Recordings

Operating System

Windows 10/11

Current Cursor Version (Menu → About Cursor → Copy)

版本: 1.1.3 (user setup)
提交: 1.96.2
日期: 979ba33804ac150108481c14e0b5cb970bda3260
Electron: 2025-06-15T06:55:04.603Z
ElectronBuildId: 34.5.1
Chromium: undefined
Node.js: 132.0.6834.210
V8: 20.19.0
OS: 13.2.152.41-electron.0

Does this stop you from using Cursor

No - Cursor works, but with this issue

i got the same issue too

I think it’s currently based on tokens, so even if you cancel, it still stays active? Not sure if that’s a bug or not.