Cursor usage explained: what counts, where to see it, what it costs a team

Cursor usage is the part of your plan's included AI allowance you have spent, plus anything charged on top from prepaid credit or on-demand billing. On current plans it is one weekly pool for Cursor's own models, visible in the editor settings and under Usage at grok.com/dashboard, while Claude, GPT and other third-party models bill separately from the first request.

For a founder paying for a few seats, that split is where surprise bills come from. Here is what counts, where to see it and how to keep a record, using Cursor's help pages as of November 2026.

What counts as Cursor usage

Cursor now sells its plans as part of The X Subscription, which also covers Grok and X. Three kinds of activity show up:

1. Cursor models from the pool. Grok 4.7, Grok 4.6, Grok 4.5 and Composer 2.5 draw from the weekly pool. The pool is shared across the editor Agent, the CLI, Cloud Agents, Grok Bot and the Grok apps, so time spent in Grok counts too.
2. Third-party models from credit. On Premium and above, Claude, GPT and other providers' models are paid from prepaid credit at API list price. They never touch the pool.
3. Your own API key. On individual plans the provider bills you directly, and those requests count against neither the pool nor your credit. On Teams and Enterprise, own-key requests still carry Cursor's token rate of $0.25 per million tokens when the model includes it.

Model choice changes how fast the pool drains, and unused allowance does not roll over.

Where to see it

When a seat hits its limit, Cursor shows a notification in the editor. On current plans the person can keep going on prepaid credit, wait for the weekly reset or upgrade.

Where spend leaks on a small team

• Claude-heavy engineers. If your team prefers Claude models, every request comes out of credit. That cost scales with usage, not seats.
• Auto Top-up. It is off until someone turns it on. Once on, it charges the saved card whenever the balance drops below a threshold. On Teams, only admins can change it.
• Caps that lag. Cursor offers a monthly additional usage limit, but says enforcement is not instant, so usage can briefly pass it.

If your engineers mostly run Claude models, compare against Claude's Team plan, which includes Claude Code from $20 a seat a month on annual billing. Cursor still wins for teams that want one editor with completions and are fine on Cursor's own models.

Keep a record per developer

Cursor's dashboards show where you stand now. To see trends per person, ask each developer to run UsageAtlas. The free desktop app reads the signed-in Cursor session, shows plan, Auto + Composer and API usage with the reset date, and keeps token history by model. It does the same for Claude Code, so a mixed team sees both tools on one screen.

UsageAtlas is open source, needs no account for local tracking and does not read prompts or responses. Its cost figures are estimates. For invoices and organization-wide controls, use Cursor's Admin Dashboard.