Cursor Pricing and Rate Limits (2026): Plans, Limits & Overages Explained

Cursor's pricing in 2026 runs from a free Hobby tier to a $200/month Ultra plan, with the $20/month Pro plan covering most solo developers. Rate limits aren't a fixed request count anymore: Cursor meters usage in dollars of underlying model cost, and once you hit your plan's included allowance you either upgrade, pay overage, or drop to slower "auto" models (see Cursor's pricing page). Teams typically land on the $40/user Business plan for pooled limits and admin controls.

What are Cursor's pricing plans?

Cursor offers four individual tiers (Hobby, Pro, Pro+, Ultra) priced at $0, $20, $60, and $200 per month, plus a Business plan at $40 per user per month for teams. Each paid tier includes a monthly allowance of model usage billed at API cost, with higher tiers unlocking more included usage and priority access to frontier models (Cursor pricing).

Plan	Price	Included usage	Best for
Hobby	$0	Limited Tab completions + one-time trial of Pro limits	Trying Cursor out
Pro	$20/month	Baseline monthly model usage allowance	Solo developers
Pro+	$60/month	Roughly 3x Pro's included usage	Heavy daily users
Ultra	$200/month	Roughly 20x Pro's included usage	Power users running many agents
Business	$40/user/month	Pooled team usage + admin controls	Teams and orgs
Enterprise	Custom	Negotiated, with SSO/SCIM and audit logs	Large organizations

How does Cursor's usage-based pricing work?

Instead of counting "fast requests" like it used to, Cursor bills included usage in dollars equivalent to the API cost of whichever model you use (Claude, GPT, Gemini, etc.). Once that monthly allowance runs out, you can enable pay-as-you-go overage billing or keep working on slower, no-extra-charge "auto" model routing.

A request routed to a Claude model draws down your balance at close to Anthropic's own API rates (Anthropic's documentation), while a GPT or Codex-style request follows OpenAI's token pricing instead (OpenAI Codex pricing). That's why two developers on the same $20 Pro plan can burn through their allowance at very different speeds depending on which model they default to.

What happens when you hit Cursor's rate limit?

When your included usage runs out, Cursor doesn't cut you off entirely: it either throttles you to free "auto" model routing or, if you've opted into usage-based pricing, bills additional requests at the underlying model's API rate. Unlike Claude Code's published 5-hour and weekly reset windows, Cursor doesn't publish a fixed daily or hourly request cap, so the exact moment you "hit a limit" depends on model choice and request size.

This makes Cursor's limits harder to predict than Claude Code's, where you can at least check /usage or a dashboard against a known reset schedule. If you also use Claude Code alongside Cursor, tools like Usagebar track that side of your usage (5-hour window, weekly cap, and context) separately in the macOS menu bar.

What's included in Cursor's free Hobby plan?

The Hobby plan is $0 and gives new accounts a one-time trial period of Pro-level limits, plus a small ongoing allowance of Tab completions and limited Agent requests once the trial ends. It's built for evaluating Cursor, not for daily production coding (Cursor pricing).

After the trial, Hobby users typically see completions and agent calls throttled significantly compared to Pro, which is the point at which most developers who like the tool upgrade.

Is Cursor Pro worth it compared to Business?

Pro ($20/month) fits solo developers managing their own usage against a personal allowance. Business ($40/user/month) fits teams because it pools usage across members, adds centralized billing and admin controls, and lets an org absorb overage instead of individual developers hitting personal caps mid-sprint.

For a team of three or more, Business also typically simplifies procurement: one invoice, one admin console, and visibility into which seats are actually using their allowance versus sitting idle (Cursor documentation).

How does Cursor's pricing compare to other AI coding tools?

Cursor's $20 Pro tier sits in the same ballpark as most competing AI coding assistants, but how fast you burn through included usage varies a lot by tool and by which underlying model you default to. If you're evaluating alternatives, see how Kiro's pricing and rate limits or Antigravity's pricing and rate limits stack up, or check Devin's pricing structure if you want a more autonomous, higher-cost tier.

If Claude Code is also in the mix, it's worth understanding its separate rate limit behavior and how to check your Claude Code usage before committing to one tool over another, since the two products meter usage very differently.

Key takeaways

• Cursor has five tiers: Hobby ($0), Pro ($20/mo), Pro+ ($60/mo), Ultra ($200/mo), and Business ($40/user/mo).
• Usage is metered in dollars of underlying model API cost, not a fixed "fast request" count.
• Once included usage is exhausted, you choose between pay-as-you-go overage or slower auto-model routing.
• Business plan pools usage across a team and adds admin/billing controls that individual Pro seats don't have.
• Cursor doesn't publish a fixed reset window the way Claude Code publishes its 5-hour/weekly limits.

Sources

• Cursor Pricing
• Cursor Documentation
• Anthropic Documentation
• OpenAI Codex Pricing
• Claude Code Usage Limits (Pro/Max)