Per-agent Grok Bot usage: who worked, tokens, cache, and cost vs the weekly pool

Describe the request

Grok Bot usage reporting is a lump-sum invoice with no timesheet.

I am on SuperGrok Heavy. The weekly pool exists, but Usage & Billing in the Grok Bot app and cursor.com/dashboard/usage stay high-level: no per-agent breakdown, no turns, no tokens, no cache vs uncached, and no record of what work those hours produced.

Grok Build is in a different place. Session logs are detailed enough to show tokens, cache %, list vs discounted cost, and cost by repo, and using grok-utils can even get plan advise. That does not cover Grok Bot agents other activities. The rest of Bot work never show up there. Routing Bot work through Grok Build would only clock the code, not the rest of the crew.

I already filed this same enhancement through Grok Bot’s in-app feedback (Send Feedback to the Cursor team). There is no public ticket URL for that card, so this forum post is the watchable request.

I do not want to build a local dashboard to fill this gap, but even if I attempted the raw log information does not expose enough details. Please expose per-agent Grok Bot usage in logs and/or the dashboard: which agent did the work, turns, tokens, cache, cost against the plans weekly/monthly Heavy and other plan) pool, and some tie to the outcome of those hours, PR’s committed, docs built etc.

Until that exists, running several Grok Bot agents is like hiring a crew on a weekly draw plus open-ended overtime, then getting a bill for hours with no names and no deliverables.

Feature request for product/service

Cursor IDE

Describe the request

Grok Bot usage in Settings is one number for the whole account.

I run several bots (a chief of staff plus specialists, plus scheduled watches). I barely type. The fleet wakes on its own. Settings still shows one Usage %. I cannot tell which bot is burning the weekly pool.

Need a per-bot meter next to that same Usage row. Same weekly pool. Just split by bot.

Not asking for a new plan or a separate quota. Visibility only.

Operating System (if it applies)

MacOS

Same gap from the machine-readable side.

Grok Bot Settings already shows weekly usage % and on-demand $ spent / $ limit. There is no public probe for that: week_pct, reset timestamp, input/output tokens, or the runtime model id.

Open-source glance.conductor (to=grok-bot) has to stay JSON null until a real Cursor probe exists. Please expose an API or documented endpoint for those fields so MCP/CLI can read them honestly instead of inventing 80% or $0 from the Settings UI.

Same weekly pool. Not a second quota. Just a probe.