Grok Bot Pricing & Usage
Grok Bot has no standalone price. It comes with a paid Cursor plan or a linked SuperGrok subscription, and each plan includes a weekly usage allowance. When that runs out, it either stops or continues on on-demand billing if you've turned that on with a cap. Usage is metered in agent steps and tokens, not messages, so one long research task can cost more than a hundred quick questions.
Plans that include Grok Bot
| Route | Plan | Relative usage (per Cursor's plans page) |
|---|---|---|
| Cursor individual | Ultra | Highest weekly usage |
| Pro+ | Generous weekly usage, below Ultra | |
| Pro | Weekly usage, below Pro+ | |
| Cursor Teams | Standard / Premium | Included for every member; usage follows the seat's allowance |
| Linked subscription | SuperGrok Heavy | Highest linked usage |
| SuperGrok Plus | Generous, below Heavy | |
| SuperGrok | Below Plus | |
| X Premium+ | Below SuperGrok Plus |
Neither xAI nor Cursor publishes Grok Bot usage as a number of steps or tokens, only these relative tiers. At the August 2026 beta launch, access was limited to SuperGrok Heavy and Cursor Ultra / Teams Premium. It has since widened to the plans above. Check your plan's current price on cursor.com or x.ai before buying.
How usage works
- Weekly allowance that resets each week.
- Metered by agent steps and tokens. Browser-heavy work (many screenshots and clicks) and long research use more than connector calls.
- Runs out → if on-demand is off, Grok Bot stops and says you've "reached your Grok Bot usage limit". If on-demand is on, work continues at model and token cost, up to the monthly spending cap you set.
- Routines count too. A routine running every 30 minutes uses your allowance around the clock. Space out schedules, or use event triggers (Slack or GitHub) so Bots only wake when there's work.
The model underneath, and API rates for comparison
xAI's current flagship is Grok 4.7, released 2026-09-21 with a 500K context. xAI says it was trained to natively understand the Grok Bot harness, which it credits for better conversational and general knowledge work. Grok Bot's own docs still don't name the model a Bot runs on, so treat that as "Grok Bot is the harness Grok 4.7 was tuned for", not as a documented model switch. If you build your own agent on the xAI API instead, these are the rates — unchanged from Grok 4.6:
| Grok 4.7 and 4.6 (API) | Input / MTok | Cached input / MTok | Output / MTok |
|---|---|---|---|
| Prompt under 200K tokens | $2.00 | $0.50 | $6.00 |
| Prompt 200K+ tokens (whole request) | $4.00 | $1.00 | $12.00 |
At 200K tokens the entire request re-bills at the higher rate. xAI also serves a fast Grok 4.7 variant at twice the output speed for twice the price. Model IDs: grok-4.7 and grok-4.6. Compare against Claude, GPT and Gemini in our cost calculator.
Ways to spend less
- Use connectors instead of the browser wherever one exists.
- Give each Bot one job. Vague Bots explore, and exploring costs steps.
- Put stopping rules in skills ("if the data isn't there after two sources, stop and report").
- Use event triggers rather than frequent schedules.
- Set an on-demand monthly cap before enabling on-demand at all.
Hub · Setup · Skills & routines · Plugins · Pricing · Security · Teams · vs OpenClaw & Hermes · FAQ · Grok Bot news
Sources: xAI's Grok Bot documentation (docs.x.ai) and Cursor's Grok Bot plans page, checked 2026-09-19. Grok Bot is in beta and changes weekly. Where the docs and a video disagree, we follow the docs and say so.