Billing of Auto and Composer 1.5 Usage in Paid Plans

Support Bug Reports
,
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

On the paid plan, Cursor allocates two separate usage buckets: “Auto + Composer” (for Auto mode and Composer model usage) and “API” (for direct API calls using other models like Sonnet or Opus). These are meant to be independent allocations. However, usage from Auto mode and the Composer 1.5 model is being incorrectly charged against the “API” bucket rather than the “Auto + Composer” bucket, which means I effectively lose my dedicated Auto+Composer allowance and instead burn through my more valuable API quota. Can you fix this?

Steps to Reproduce

Have an active paid Cursor plan with both the “Auto + Composer” and “API” usage buckets available

  1. Use Cursor in Auto mode or explicitly select the Composer 1.5 model for a task
  2. Navigate to the usage dashboard
  3. Observe that the usage counter for “Auto + Composer” remains at 0% while the “API” bucket increases

Expected Behavior

Usage from Auto mode and the Composer 1.5 model should be reflected in the “Auto + Composer” bucket. The “API” bucket should only increase when using other models (e.g., Sonnet 4.6, Opus) that are explicitly billed against API quota.

Screenshots / Screen Recordings

Bildschirmfoto 2026-03-04 um 10.23.54.png
Bildschirmfoto 2026-03-04 um 10.23.54.png1984×596 19 KB

Operating System

MacOS

Version Information

Latest IDE Version

For AI issues: which model did you use?

Auto and Composer 1.5

Does this stop you from using Cursor

Yes - Cursor is unusable

6

Some of these usage meters are heavily delayed. Is it possible you were seeing the API meter increase from past requests and the Auto/Composer hadn’t updated yet or was negligible?

Otherwise, yea this seems like a bug.

7

Hey, thanks for the report. The screenshot clearly shows 0% Auto+Composer and 48% API, which shouldn’t happen if you’re only using Auto and Composer 1.5.

First, please check this. Go to https://cursor.com/dashboard?tab=usage and share a screenshot of the per-request breakdown. It lists each individual request, the model used, and the cost. This will help confirm that Auto and Composer 1.5 requests are really being counted under the wrong category, and not that Auto quietly picked a different model.

If the numbers confirm it, I’ll pass this issue to the team.