Quick answer
Cursor documents included agent usage by plan as Pro: extended Agent limits; Pro+: 3× Pro; Ultra: 20× Pro (no dollar amount published). Included usage is split into two monthly pools—Cursor models and third-party models charged at provider prices—and on-demand usage continues at API rates once a pool is exhausted.
Verified Sep 17, 2026Official source
Current limits
| Constraint | Current value | Scope | Verified source |
|---|---|---|---|
| included Agent usage by planIncluded usage is split into two monthly pools—Cursor models and third-party models charged at provider prices—and on-demand usage continues at API rates once a pool is exhausted. | Pro: extended Agent limits; Pro+: 3× Pro; Ultra: 20× Pro (no dollar amount published) | Individual paid plans (Pro $20, Pro+ $60, Ultra $200 per month) | CursorSep 17, 2026 |
Why does this limit matter?
Two users with the same plan can reach the allowance at different times because models have different provider costs.
This value is scoped to Individual paid plans (Pro $20, Pro+ $60, Ultra $200 per month); a different plan, runtime, model, endpoint, region, or account can produce a different effective constraint.
What should you check?
- Review the Cursor usage dashboard and the model used by expensive agent sessions.
- Confirm the exact plan, model, runtime, endpoint, region, and account that serve the failing workload.
- Record the observed value, response headers or configuration, timestamp, and source without logging secrets.
Important caveats
- Cursor no longer expresses included usage as a dollar amount; plan multipliers describe Agent limits relative to Pro, not a guaranteed request count.
- Treat the official source and live account configuration as authoritative if they differ from this verified snapshot.
HyperObserve reports the documented platform constraint. Your application, SDK, gateway, provider, region, or account can impose a lower effective limit.
Related