Key takeaways
- Cursor Pro has been priced at $20/month since 2023. The sticker never changed — but the June 2025 credit switch cut heavy frontier-model users' effective requests by ~55%, prompting a CEO apology and refunds.
- Three key changes: ①March 2023 fixed-request quota; ②June 16, 2025 API credit model (Pro+ $60 / Ultra $200 added); ③August 2025 Teams moved to variable billing.
- The 2026 four-plan lineup is stable across 10+ price snapshots. But Auto mode credit policy can shift without notice.
- Next most likely change: Teams/Enterprise increase ($40→$50), then Pro toward $25 in 2027–2028.
- The real dividing line is not "will it raise prices" but which credit band your usage falls into — Auto mode users are largely insulated; heavy manual Frontier users carry the highest risk.
Bottom line: the label stayed at $20, the contents changed
Sticker price is not the dividing line — billing model (quota vs credit) and your monthly usage band are.
As of August 2026, Cursor Pro is still $20/month. But what that $20 covers has been structurally rebuilt three times. The sharper question is not "will Cursor raise prices" but: does your usage land in the Auto mode band or the manual Frontier model band? Keywords: Cursor Pricing · Cursor Price · Cursor Subscription
1. Why Cursor adjusts pricing
Cursor's model: charge users a fixed subscription, pay actual API costs to Anthropic / OpenAI / Google. Three variables shifted simultaneously by 2025:
- Heavy users multiplied: Agent workflows spread; per-user daily requests grew from tens to hundreds.
- Model cost volatility: Claude Opus costs ~$0.08–0.12/request; at 500 requests, Cursor's API bill per Pro seat approached $20.
- Model breadth expanded: GPT-5.6, Gemini Pro added; cost gap between cheapest and priciest model up to 25×.
A flat quota became structurally loss-making for heavy frontier users. Credit-based billing was inevitable — the same logic that drove GitHub Copilot's metered shift in June 2026.
2. Change 1 — March 2023: Fixed-request quota launch
- Hobby (free): limited trial, no Agent.
- Pro ($20/month): 500 fast requests/month (premium models) + unlimited slow requests.
Community reception was overwhelmingly positive. Reddit and Hacker News praised it as best value in AI coding tools. This honeymoon lasted until June 2025.
3. Change 2 — June 16, 2025: API credit switch (highest controversy)
Pro remains $20/month, but quota switches from "500 fast requests" to "$20 in API credits":
- With Claude Opus (~$0.088/request): ~225 requests — a ~55% reduction.
- Credits depleted → overage billing activates (can be capped).
- Simultaneously added: Pro+ ($60/month, ~$70 credits) and Ultra ($200/month, ~$400 credits).
The switch happened with minimal warning; many users found overage charges only at month-end. For a parallel look at another AI coding tool: Claude Code 2026 Pricing Plans Explained.
4. Change 3 — August 2025: Teams variable billing
Teams ($40/user/month) moved from fixed per-request costs to the same variable API billing as individual plans:
- Each seat gets Pro-equivalent credits, depleted at actual API rates.
- Auto mode adjusted to discounted-first on all paid plans.
- Teams admins can set per-seat overage caps centrally.
Reception was calm — after June's turmoil, users had already adjusted. See Cursor Background Agents persistent Cloud Mac setup.
5. Three-change comparison table
| Dimension | Change 1 (Mar 2023) | Change 2 (Jun 2025) | Change 3 (Aug 2025) |
|---|---|---|---|
| What changed | Paid plans launched, fixed quota | Switched to API credits | Teams moved to variable billing |
| Pro price | $20 | $20 (unchanged) | $20 (unchanged) |
| Pro included | 500 fast requests | $20 API credits | $20 credits (unchanged) |
| New tiers | None | Pro+ $60, Ultra $200 | None |
| Rollout | Public launch, welcomed | Silent switch, surprise overages | Advance notice, calm |
| User reaction | Positive | Strong backlash, CEO apology + refunds | Generally accepted |
| Who was impacted | All users | Heavy frontier-model users most | Teams admins |
6. Current plans five-axis comparison (August 2026)
| Plan | Price | Credit pool | Execution | Permission | Best for |
|---|---|---|---|---|---|
| Hobby | $0 | Limited Agent | Basic completion | No agent exec | Trial / student |
| Pro | $20/mo | $20 credits (~$16 annual) | Full Agent + Tab | Basic MCP | Daily personal dev |
| Pro+ | $60/mo | ~$70 credits (3×) | Full Agent + Tab | Full MCP | 4–6h/day heavy dev |
| Ultra | $200/mo | ~$400 credits (20×) | Priority model access | Full + priority | Full-time Agent workflow |
| Teams | $40/user/mo | Pro-equiv per seat | Full + SSO | Central billing | 5+ engineering team |
| Enterprise | Custom | Pooled credits | Full | Compliance + SCIM | 50+ org |
Auto mode note: Auto is discounted by Cursor's scheduler. Manually choosing Claude Sonnet/Opus/GPT-5.6 burns credits at full API rates.
7. Scenario decision matrix
| Use case | Recommended plan | Key reason | Monthly cost |
|---|---|---|---|
| Occasional / learning project | Hobby | Free, covers basics | $0 |
| 1–2h/day IDE, mostly Auto | Pro | $20 credits sufficient | $20 |
| 4h+/day, Agent tasks | Pro+ | 3× credits, no month-end anxiety | $60 |
| Full-time + parallel worktrees | Ultra | 20× credits, priority access | $200 |
| 5+ engineering team | Teams | Unified billing + SSO | $200+/mo (5 seats) |
| CI/CD / headless Agent | API pay-as-you-go | No monthly fee | Per usage |
8. Recommended stacks
Light individual: Cursor Pro $20/mo (annual ~$16) + Auto mode only → bill locked at $20 Heavy individual: Cursor Ultra $200/mo + Cloud Mac mini M4 (kvmboot daily/monthly) Team: Cursor Teams $40/user/mo + overage cap at $0
9. Common mistakes
- Mistake 1: "$20/month unchanged = no price increase." Sticker stayed, but effective requests for heavy frontier users dropped ~55% in June 2025. It is a de facto price increase.
- Mistake 2: "Ultra means no overages." Ultra's ~$400 credit pool can still bottom out with parallel Opus sessions. Always set an overage cap.
- Mistake 3: "Auto and manual mode cost the same." Auto is discounted by Cursor. Manual Opus/Sonnet burns credits at full API rates — often 10–20× more.
- Mistake 4: "Annual 20% discount is permanent." Discount rates can change. Verify the current rate before committing to a year.
- Mistake 5: "Pro is enough, I don't need to upgrade." Frequently picking Sonnet/Opus manually? $20 credits run out by day 15. Pro+ is cheaper than buying overages.
- Mistake 6: "Price changes only affect individuals." The August 2025 Teams change showed enterprise plans are equally exposed. Admins must configure caps from day one.
10. 7-step checklist
- Audit two weeks of usage: Check monthly usage report; note Agent hours vs Tab completions; Auto vs manual model split.
- Identify your primary model: Mostly Auto? Pro usually fine. Frequent Frontier manual picks? Calculate whether monthly credit burn exceeds $20.
- Set overage cap to $0 immediately: Regardless of plan, cap overages first.
- Run 30 days on Pro: Complete a real project; note which day credit warning fires.
- Upgrade by ROI: Credits depleted before day 20? Weigh $40 delta for Pro+.
- Optimize execution environment: Move long Agent jobs to Cloud Mac (persistent tmux, no sleep interruption). See Cursor Background Agents cloud setup.
- Monthly TCO review: Combine subscription + overages + cloud execution; adjust plan with data.
11. 2026 price forecast
| Change type | Likelihood | Timeline | Rationale |
|---|---|---|---|
| Teams/Enterprise increase ($40→$50) | High | Q4 2026–Q1 2027 | Enterprise growth, SSO/SCIM feature cost |
| Pro increases from $20 to $25 | Medium | 2027–2028 | Following GitHub Copilot's metered shift |
| Annual discount narrows (20%→15%) | Medium | 2026–2027 | Lock-in effect diminishes; recapture margin |
| Auto mode credit policy tightens | Medium | Any time | Auto scheduling cost grows with volume |
| Hobby plan removed / restricted | Low | — | Hobby is acquisition funnel |
Asymmetric conclusion: Cursor will not easily touch the $20 Pro sticker — it's a brand anchor. But credit pool rules, Auto discounts, and annual terms can shift silently at any time, with impact equivalent to an overt increase for heavy users.
For a look at how other AI coding tool costs stack up: GPT-5.6 for coding — Sol, Terra, Luna vs Claude and Gemini decision guide.
12. FAQ
Did Cursor raise prices?
The sticker hasn't changed (Pro still $20/month). The June 2025 credit switch cut effective requests for heavy frontier users by ~55%; CEO apologized and issued refunds. For heavy users, it was a de facto price increase.
How many requests does Pro $20/month buy now?
Depends on model: Auto mode covers a full month of normal dev. Manual Claude Opus (~$0.088): ~225 requests. Manual Haiku (~$0.002): ~10,000 requests.
Pro+ $60 vs Ultra $200 — which to pick?
Credits depleted before day 20 → upgrade to Pro+ (3×, $60). Pro+ also bottoms out with parallel worktrees / long Opus runs → upgrade to Ultra (20×, $200). Both support monthly billing.
Will Teams pricing change?
Teams ($40/user/month) is the most likely candidate for the next increase. Enterprise pricing adjustments are historically far better absorbed than consumer-tier increases.
How do I minimize Cursor costs?
①Default to Auto mode; ②set overage cap to $0; ③use annual billing (~20% off); ④move long Agent sessions to Cloud Mac to eliminate sleep-retry credit waste.
13. Summary
Will Cursor raise prices? The number is stable for now. Three changes show a clear trajectory: from "fixed service" to "metered elasticity" — where the whole AI tools industry is heading, Cursor completed the migration earlier than most. For Auto-mode users with under 2h/day IDE usage, Pro at $20 remains one of the best-value AI coding subscriptions. Heavy Agent users are honestly served by Pro+ or Ultra.
Cursor Agent on Cloud Mac: credits go toward delivery, not sleep-retry waste
Cursor Ultra at $200/month buys credit capacity — but every laptop sleep breaks an Agent session and retry consumes credits silently. Move long Cursor Agent jobs to kvmboot Cloud Mac mini M4: persistent tmux sessions, worktree isolation, real Apple Silicon execution, macOS Gatekeeper + SIP reducing script risk. Every retry eliminated is real credit savings.
Start with a daily rental to validate your Cursor Agent workflow, then move to monthly — kvmboot Cloud Mac mini M4 is the shortest path to full value from your Cursor subscription, see plans and put every credit toward shipping.