Read sand-* and grok-bot-* tokens next to Cursor Other Models

Cursor’s Spending page often lists Grok Bot traffic as sand-default, sand-automation, or later labels such as grok-bot-default, grok-bot-automation, and grok-bot-cua. Those rows can sit under Other Models or Cursor Models while a separate Grok Bot weekly bar also exists. Staff answers evolved. This page quotes them in order and stops where the matrix is still open.

Staff map (use this first)

On the sand-* tokens thread, Cursor staff (mohitjain) corrected the story on Sep 3, 2026:

1. Conversations draw from the Grok Bot weekly pool and do not touch Other Models.
2. Routines / automations and browser actions (anything Grok Bot runs as a cloud agent) run on your Cursor plan, so they count against Other Models and can block Claude/GPT even while the Grok Bot weekly pool still has room.
3. That cloud-agent portion is not display-only.
4. To unblock Claude/GPT now: pause or delete heavy routines/automations (grok-bot-automation / grok-bot-cua are the main drain), or enable paid on-demand, or wait for quota reset. Account review: hi@cursor.com.

Earlier on the same thread (Aug 27), staff had said sand-default / sand-automation were weekly-only and display-grouped under Other Models. Treat the Sep 3 reply as the current staff map for Pro/Ultra-style Other Models burn.

Teams / Cursor Models rows (display)

On the Teams Cursor Models thread, staff (mohitjain) wrote that grok-bot-* rows under Cursor Models are grouped by the model that served the request, not by product; those rows are display-only for the Cursor Models %, and Grok Bot still has its own weekly pool in the Bot app. Staff also repeated: cloud agents launched from Grok Bot do count toward your Cursor plan.

Still open on the dual-pool matrix

Forum 170951 asks for a per-row matrix (grok-bot-default / grok-bot-automation / grok-bot-cua) and whether one task can burn both pools. As of the watch that logged it, that thread had no staff reply. Do not invent the cell-by-cell table. Cite Sep 3 + Teams display notes above, then escalate with screenshots if your meters disagree.

Read both meters side by side

1. Open Grok Bot Usage & Billing (or the in-app weekly bar) and note the percentage and reset time.
2. Open the Cursor Spending / Included Usage dashboard for the same account.
3. Find rows labeled sand-* or grok-bot-* under Other Models and/or Cursor Models.
4. List active routines and any recent browser / computer-use / Cloud Agent launches from Bot.
5. Note whether On-Demand is enabled.
6. If Claude/GPT are blocked while Bot weekly still has room, pause heavy grok-bot-automation / grok-bot-cua work first (per Sep 3 staff).

Related: Understand Grok Bot usage and “Bot failed to respond”, Control Grok Bot On-Demand after weekly usage hits 100%, Spend your Grok weekly usage pool.

Pitfalls

• Bot FAQ eligibility (SuperGrok Plus / Heavy, Cursor Pro+ / Ultra / Teams Standard or Premium) still governs who gets Bot. Dual Cursor + SuperGrok uses whichever has more usage.
• A coding agent on cursor.com/agents is Cursor plan spend, per staff.
• METAGROK will not invent a billing fix or a finished dual-pool matrix. Quote staff dates, quote your meters, and file with Cursor when the UI and the Sep 3 map disagree.