Cursor Usage Limit Explained: What Happens When You Run Out of Included Usage
Cursor doesn't count discrete requests anymore — it meters two dollar-denominated usage pools that reset with your billing cycle. Here's how Cursor's Pro/Pro Plus/Ultra quotas actually work, what happens when you run out, and how that compares to Claude Code, Codex, and Grok's time-based limits.
Claude Code, Codex, and Grok all hit you with the same shape of wall: a rolling time window that clears at a specific clock time. Cursor doesn't work that way at all, and that mismatch is exactly why "how much Cursor usage do I have left" is a confusing question to search for. Cursor doesn't meter time — it meters dollars of model usage against a monthly pool, and what happens when the pool runs dry depends on a setting you may not know you have.
Cursor doesn't count requests anymore
Older Cursor plans used to advertise a flat number — "500 fast requests a month" — which made the limit easy to reason about even if it was easy to blow through. That request-counting model is now legacy. Current plans meter two separate usage pools, each priced in dollars and reset on your monthly billing cycle:
- Cursor Models pool — usage of Cursor's own-hosted models (Grok 4.6, Grok 4.5, Composer 2.5), priced per token but bundled generously into every paid plan.
- Other Models pool — usage of third-party models (Claude, GPT, Gemini, and the rest), billed at that model's real API rate out of the same included dollar allowance.
Both pools show up in your editor settings and usage dashboard as a running dollar total, not a request counter. Since different models cost different amounts per token, which model you pick — and whether you use Fast mode — directly changes how fast your included usage burns down.
Connect the Claude or Codex you already pay for — the rest runs on workers that cost a fraction.
Download meshcode →Plans and what they include
| Plan | Price | Included usage |
|---|---|---|
| Hobby | Free | Limited Agent requests, no credit card |
| Pro | $20/mo | Both pools included, extended Agent limits |
| Pro Plus | $60/mo | 3x Pro limits on Agent |
| Ultra | $200/mo | 20x Pro limits on Agent |
| Teams Standard | $40/user/mo | Same as Individual, plus team admin |
| Teams Premium | $120/user/mo | 5x Standard limits |
As a rough sizing guide from Cursor's own docs: daily Tab-only users and light Agent users typically stay inside their included pool; daily Agent users usually land around $60–100/month in total usage; power users running multiple agents or automations often exceed $200/month. If those numbers already sound closer to a metered bill than a flat subscription, that's the point — Cursor's pricing behaves like usage-based billing with a subscription-shaped floor, not the other way around.
What actually happens when you hit the limit
This is the part that surprises people coming from Claude Code, Codex, or Grok, where hitting the limit means a hard stop until a specific reset time. Cursor's docs are explicit: on-demand usage bills at the same API rates and never downgrades quality or speed. In practice, once your included pool is spent, you have two options:
- Let on-demand billing continue — same model, same speed, billed at list price for whatever you use past the included amount.
- Upgrade your plan — move to a higher tier for more included usage instead of paying overage rates.
There's no forced wait, no quality throttle, no "come back at 9:48 PM" banner — as long as on-demand usage is enabled on your account. If you've turned that off to cap spend, running out behaves more like the other tools: agent requests stop until the pool resets with your next billing cycle.
One extra line item to know about: on Teams and Enterprise plans, third-party model requests carry a Cursor Token Rate of $0.25 per million tokens on top of that model's own API price. First-party Cursor models (Grok, Composer) are exempt.
Why this is a different mental model, not just different numbers
Claude Code's "resets at 5PM," Codex's "try again at Jul 20th," and Grok's "20 questions per 2 hours" are all answers to when can I use this again. Cursor's dashboard answers a different question: how much have I spent, and what's left. Neither shape is objectively better — a hard time wall caps your worst-case bill for you automatically; a dollar pool with on-demand billing keeps you working but leaves the spending cap up to you to configure.
The meshcode angle
meshcode is a native desktop app for macOS and Windows built around multiple agent panes running side by side. You can connect your existing Cursor-compatible model access or your own Claude Code/Codex CLI into a pane with no extra token charge from meshcode, and put meshcode's own pay-as-you-go mesh models in another pane — a metered balance you top up from $1, with no monthly minimum and no dollar pool to track across two separate categories. For the rest of this series, see what Claude Code's usage limit message actually means, what Codex CLI's banner is really telling you, and Grok's different limit shape.
👉 Download meshcode — Mac, Windows