Eight months ago our dev team ditched VS Code + GitHub Copilot and moved to Cursor. The short version: Cursor Pro at $20/month replaced a $19/month Copilot subscription AND cut our debugging time by roughly 40%. Since then Cursor has overhauled how its plans meter usage, so here is the current pricing -- verified against cursor.com/pricing on August 26, 2026 -- plus the non-obvious gotchas.
Cursor Plans (August 2026)
Hobby: Free
- Limited model usage -- enough to evaluate the editor, not to code on daily
- Tab autocomplete included
- All IDE features: multi-file editing, codebase search, terminal integration
Honest take: the free allowance sounds generous until you realize how constantly the AI fires during active coding -- every pause between keystrokes can trigger a completion. In our experience a few days of real work exhausts it. Hobby is a trial, not a plan you can work on.
Pro: $20/month
- A set amount of model usage per month (Cursor no longer publishes a request count)
- Access to premium models (GPT-5.6 Sol, Opus 5, Fable 5)
- Agent mode, Composer multi-file edits, Cmd+K inline edits
- Optional usage-based billing once you exceed the included allowance
Honest take: Pro is the plan. The included allowance covers typical daily development for most of our team; the members who lean hard on agent mode are the ones who hit it. When that happens you either enable usage-based billing (pay for what you burn) or move up a tier.
Pro+: $60/month and Ultra: $200/month
- Same features as Pro with progressively larger usage allowances
- Aimed at developers running long agent sessions or heavy multi-file generation all day
Honest take: these tiers exist because agent-style coding consumes model usage at a completely different rate than chat and autocomplete. If your usage-based overage on Pro consistently exceeds $40/month, Pro+ is the cheaper way to buy the same headroom; the same math at the top end points to Ultra.
Teams: $40/user/month
- Everything in Pro
- Centralized team billing and seat management
- Usage analytics per team member
- SSO/SAML (OpenID Connect)
- Admin dashboard with policy controls
- Privacy mode enforcement (code never stored on Cursor servers)
- Priority support
Honest take: The 2x price jump from Pro to Teams is steep. But for teams, the centralized billing alone justifies it -- no more expensing 8 individual $20 subscriptions. The privacy mode enforcement matters if you work with client code or proprietary IP. Without Teams, individual devs can toggle privacy mode off (and forget to turn it back on). Enterprise (custom pricing) sits above it for larger orgs.
How Usage Metering Works Now
This is the most confusing part of Cursor pricing -- and it changed. Earlier plans counted "fast premium requests" (Pro included 500/month), and most third-party articles still describe that system. Cursor has retired it. What the pricing page says now:
- Each paid plan includes "a set amount of model usage." Cursor does not publish a fixed request count; heavier models and longer agent runs consume the allowance faster.
- Usage-based billing is optional beyond that. You can let the plan hard-stop, or pay for additional usage at metered rates.
- Tab completions remain the cheap layer; chat, Composer, and agent runs on frontier models are what actually drain the allowance.
Practical consequence: you can no longer budget by counting requests. Watch the usage dashboard for your first two weeks, see where your team actually lands, and pick the tier (or usage-billing cap) from data.
Cursor vs GitHub Copilot: Price Comparison
- Copilot Pro: $10/month -- cheapest paid AI coding option (Pro+ is $39, and a new Max tier runs $100)
- Copilot Business: $19/user/month -- team features, policy controls
- Copilot Enterprise: $39/user/month -- codebase-aware completions
- Cursor Pro: $20/month -- more capable but slightly pricier than Copilot Pro
- Cursor Teams: $40/user/month -- comparable to Copilot Enterprise
Copilot is cheaper at every tier. But Cursor is more capable -- especially for multi-file edits, codebase-aware reasoning, and complex refactoring. The question is whether that capability gap is worth $10/month more per developer. For our team, yes -- the time saved on debugging and refactoring pays for the difference many times over.
Read our full Cursor vs Claude Code comparison or GitHub Copilot vs Cursor comparison for detailed feature breakdowns.
Hidden Costs
- Usage-based overage: the successor to the old API-key trap. Monitor per-seat usage the first month -- a single codebase-wide agent task can consume a surprising share of a month's allowance in one sitting.
- Learning curve: Cursor has a 1-2 week productivity dip while your team learns the new workflows (Composer, Cmd+K patterns, .cursorrules files). Factor in the lost productivity.
- Extension compatibility: Most VS Code extensions work, but some do not. If your team relies on niche extensions, test compatibility before committing to annual billing.
Our Recommendation
- Solo developer, budget-conscious: Start with Hobby to test, then upgrade to Pro when you hit limits (you will within a week).
- Solo developer or freelancer: Pro ($20/month). No question.
- Heavy agent-mode user: Pro with a usage cap first; move to Pro+ ($60) or Ultra ($200) only once your measured overage says so.
- Dev team (3-10 people): Teams ($40/user/month) for centralized billing and privacy controls. Or Pro with individual billing if budget is tight.
- Enterprise (50+ devs): Contact Cursor for custom pricing. Or evaluate whether GitHub Copilot Enterprise at $39/user is a better fit for your stack.
We help development teams evaluate and implement AI coding tools. Whether it is Cursor, Claude Code, Copilot, or a custom setup, we can help you pick the right tool and configure it for your codebase. Book a consultation.
Frequently Asked Questions
Is Cursor free?
Yes, the Hobby plan is free with limited model usage. It is enough to evaluate Cursor but not to use it as your daily driver.
What happened to Cursor's 500 fast requests?
Cursor retired the request-counting model. Paid plans now include a set amount of model usage per month, with optional usage-based billing beyond it -- and two new individual tiers, Pro+ ($60) and Ultra ($200), for developers who need more headroom than Pro provides.
Is Cursor better than VS Code + Copilot?
For AI-assisted coding, yes. Cursor's multi-file editing (Composer), codebase-aware chat, and inline edit (Cmd+K) are more capable than Copilot's chat and completions. For pure autocomplete, the difference is smaller. Many developers keep Copilot for completions and add Cursor for complex tasks, but that means paying for both.
Does Cursor store my code?
By default, code snippets sent to AI models may be stored temporarily. Privacy mode (available on all plans, enforced on Teams) ensures code is never stored on Cursor's servers or used for training. Enable this for any proprietary or client code.