Grok Bot token usage
Grok Bot token usage is not a per-message counter in chat. Eligible accounts track weekly included usage and optional on-demand overage through app settings or the Cursor account page.
Hub: Tokenminning in Grok Bot.
Where to read the meter
| Surface | Location | What it shows |
|---|---|---|
| Usage & Billing | Settings → Usage & Billing (when available) | Weekly included and on-demand usage |
| Weekly usage | Account menu → Weekly usage (when available) | Quick weekly snapshot |
| Cursor account | Cursor dashboard | Fallback when in-app surfaces are hidden |
| Enterprise | Organization administrator | Org-level rollups |
If neither in-app surface appears, review usage from the Cursor account page or contact your administrator.
What drives consumption
| Driver | High usage means |
|---|---|
| Long conversations | History resends every turn — new chat per outcome |
| Computer use | Browser and terminal steps on the shared VM |
| Background routines | Scheduled or event-triggered work while you are away |
| Connector breadth | Many Plugins enabled; narrow to the workflow |
| Catch-all Bots | Vague descriptions → more exploration — Bots |
| Write actions | Unscoped sends, publishes, or production edits without drafts first |
Plan eligibility
Grok Bot requires an eligible plan: SuperGrok Plus, SuperGrok Heavy, Cursor Pro+, Cursor Ultra, or Cursor Teams Standard or Premium. Sign in with your Cursor account in the Grok Bot desktop app .
Grok Bot requires cloud data storage. Accounts on Legacy Privacy Mode must update Cursor data settings before the app can start.
After a heavy week
- Weekly usage or Usage & Billing — did on-demand spike?
- Open View conversation details → Routines — any routine with frequent failures or broad triggers?
- Match the spoke:
- Input-heavy threads → Bots, new conversation per task
- Overnight burn → Routines
- Browser-heavy work → Computer & apps
- Unexpected writes → Approvals
Time zone and schedules
Routines use the Timezone under Settings → General → Agent. Wrong time zone schedules work when you are not reviewing results — leading to piled-up approvals and retry loops.