QUICK ANSWER
● Trying Cursor: Hobby (free) — limited agents and Tab, no credit card
● Daily professional use: Pro at $20/month — unlimited Tab, $20 monthly credits, Auto mode unlimited
● Hitting Pro limits regularly: Pro+ at $60/month — 3x credits ($60 pool), identical features to Pro
● Power user, all-day agent use: Ultra at $200/month — 20x credits ($400 pool), priority features access
● Team of 3+: Teams at $40/user/month — admin controls, SAML SSO, centralized billing, usage analytics
● Students: Free for 1 year of Pro with .edu email ($240 value)
Every Cursor Plan - Full Breakdown
| Plan |
Monthly price |
Annual (20% off) |
Monthly credit pool |
Tab completions |
| Hobby |
Free |
— |
Limited |
Limited |
| Pro |
$20/month |
$16/month ($192/yr) |
$20 |
Unlimited |
| Pro+ |
$60/month |
$48/month ($576/yr) |
$70 (3.5x) |
Unlimited |
| Ultra |
$200/month |
$160/month ($1,920/yr) |
$400 (20x) |
Unlimited |
| Teams |
$40/user/month |
$32/user/month |
Pro-equivalent per seat |
Unlimited per seat |
How the Credit System Actually Works
This is where most developers get confused. Cursor has two billing modes that run simultaneously:
Auto mode — Unlimited on all paid plans
When you use Cursor in Auto mode (the default), it selects the optimal model for your task automatically. Auto mode does NOT consume your credit pool. It is effectively unlimited on Pro, Pro+, Ultra, and Teams plans. If you use Cursor primarily in Auto mode, your $20 credit pool on Pro may last much longer than you expect.
Premium model requests — Draw from your credit pool
When you manually select a specific premium model (Claude Opus 4.8, GPT-5.5, Gemini 2.5 Pro), each request draws from your monthly credit pool. Roughly: 225 Claude Sonnet 4.5 requests, 550 Gemini requests, or 500 GPT-5 requests per $20 Pro credit pool. Agent Mode multiplies this — each background step counts as a separate model call.
Max Mode — The expensive one
Max Mode expands the context window beyond the 200K default (to 1M tokens for Gemini 2.5 Pro and GPT-4.1). This consumes significantly more credits per request. Only enable Max Mode when you genuinely need the extended context — for most files, 200K is sufficient and Max Mode burns through your credit pool fast.
What Each Plan Gets You in Practice
Hobby - Free Forever
The Hobby plan is free with no credit card required and does not expire. You get a limited number of Agent requests and Tab completions per month — enough to genuinely evaluate Cursor, not enough for professional daily use. Students with a .edu email get a full year of Pro free ($240 value). If you write code professionally, start with the 7-day Pro trial before committing.
Pro - $20/Month (The Right Starting Point)
Pro at $20/month is the right plan for most working developers. Unlimited Tab completions mean the autocomplete that most developers use all day costs nothing beyond the subscription. Unlimited Auto mode means you can run Agent tasks in Auto without consuming credits. The $20 monthly credit pool is for premium model requests — 225 Sonnet requests or 500 GPT-5 requests per month. For developers who use Auto mode primarily with occasional premium model requests, Pro's credit pool covers the entire month comfortably. The signal to upgrade: if you're rationing agent requests in the last week of the month or regularly paying overages above $20.
Pro+ - $60/Month (The Recommended Upgrade)
Pro+ is identical to Pro in features. The only difference is 3x the credit pool ($70 per month vs $20). Cursor itself labels Pro+ as the "recommended" tier. The math: if you are regularly hitting Pro's credit limit and paying $20-40 in monthly overages, Pro+ at $60 costs the same or less while eliminating the usage anxiety. Upgrade to Pro+ if you are spending 4+ hours daily in Cursor running multi-file edits, agent tasks, and complex refactors with premium models.
Ultra - $200/Month (Only If You Live in Cursor)
Ultra gives 20x Pro credits ($400/month pool) plus priority access to new features. At $200/month, this is infrastructure spend rather than a productivity subscription — it is built for full-time AI-native developers running Background Agents continuously, working on large codebases with extensive context, and relying on frontier models for nearly every task. Most developers will never need Ultra. The signal you need it: regularly exhausting Pro+ credits before the end of the month despite using Auto mode efficiently.
Teams - $40/User/Month
Teams ($40/seat/month) adds organizational controls on top of Pro-equivalent AI access: shared team rules and prompts, centralized billing, usage analytics per developer, org-wide privacy mode controls, role-based access control (RBAC), and SAML/OIDC SSO. The AI capability per seat mirrors Pro — each developer gets unlimited Tab and Auto mode plus the $20 credit pool. For teams of 3+ who need IT/security compliance, Teams is the right tier. For teams that only want more AI power without admin features, individual Pro subscriptions at $20/seat are half the price.
The 4 Things That Burn Credits Fastest
1. Manually selecting frontier models — every time you pick Claude Opus 4.8, GPT-5.5, or Gemini 2.5 Pro instead of using Auto mode, you draw from your credit pool. Auto mode is unlimited; manual premium model selection is not.
2. Agent Mode with premium models — Agent Mode runs multiple model calls per task. Each background step counts separately. A complex agent task on Claude Opus can cost 10-20 regular requests worth of credits in one session.
3. Max Mode — enabling Max Mode expands context to 1M tokens but consumes credits significantly faster per request. Only enable when you genuinely need context beyond 200K tokens.
4. Large codebase context — when Cursor reads many files to build context for an agent task, every token in those files counts toward the request cost. Larger codebases cost more per agent request even at the same task complexity.
How to Make Pro Credits Last Longer
Use Auto mode by default — Cursor's Auto mode is unlimited on all paid plans. Let Cursor pick the model for most tasks. Only manually select a premium model when the task specifically requires frontier model quality.
Use Tab completions for routine work — Tab is unlimited on all paid plans. For code completion, small edits, and routine suggestions, Tab does not touch your credit pool. Reserve Agent mode for tasks that genuinely need multi-step reasoning.
Annual billing saves 20% — on Pro that is $48/year in savings ($192 vs $240 annually). Commit after one month of confirmed daily use.
Track Pro+ overage math before upgrading — if you are paying $20-40/month in overages on Pro, Pro+ at $60 saves money. If you are paying less than $20 in overages, stay on Pro with overages — it is cheaper than Pro+.
Cursor vs Claude Code Pricing - Side by Side
| Use case |
Cursor plan |
Claude Code plan |
Which wins on cost |
| Solo dev, light daily use |
Pro $20 |
Pro $20 |
Tie — both $20 |
| Power user, heavy premium models |
Pro+ $60 |
Max 5x $100 |
Cursor — $40 cheaper |
| Multi-agent, all-day use |
Ultra $200 |
Max 20x $200 |
Tie — both $200 |
| Teams (3-10 developers) |
Teams $40/seat |
Team Premium $100/seat |
Cursor — $60/seat cheaper |
| CI/CD heavy pipelines |
N/A (no API) |
API $3-25/M |
Claude Code (only option) |
Frequently Asked Questions
Is Cursor free?
Yes — the Hobby plan is free forever with no credit card required. It has limited Agent requests and Tab completions, enough to evaluate Cursor but not for daily professional use. Students with a .edu email get one year of Pro free. A 7-day Pro trial is also available for non-student professionals.
Is there a difference between Pro+ and Ultra features?
Pro+ and Ultra have identical features to Pro. The only differences are the credit pool size (Pro+ 3x at $70/month, Ultra 20x at $400/month) and Ultra's priority access to new features. There are no capability gates between tiers — you do not get new tools by upgrading, only more usage headroom.
Why do credits run out before the end of the month?
Three causes: manually selecting premium models instead of Auto mode, running Agent Mode with premium models (each background step = separate credit consumption), or enabling Max Mode which expands context to 1M tokens at higher cost. Switch to Auto mode for routine tasks — it is unlimited — and reserve manual premium model selection for tasks where frontier model quality is genuinely required.
Should I choose Cursor or Claude Code?
They solve different problems and most serious developers use both. Cursor for daily IDE coding — best Tab autocomplete, Background Agents inside VS Code, fast inline editing. Claude Code for long autonomous sessions — 12+ hours unattended, 3,000+ MCP integrations, CI/CD pipelines. If you can only pick one: start with Cursor for daily workflow, add Claude Code when you need overnight agentic tasks. See our four-way comparison for the full breakdown.
Related: Claude Code Pricing Explained · Cursor vs Claude Code vs Codex vs Windsurf · Best AI Coding Agents 2026