🇺🇸 USA · Cursor
Status: 🟩 COMPLETE 🟦 LIVING Last updated: 2026-06-26 Plain-English tagline: The AI-first code editor — a fork of VS Code rebuilt around always-on AI assistance. The most-used Claude Code alternative; for many developers, the “what coding feels like in 2026” benchmark.
Front-matter facts
| Field | Value |
|---|---|
| Vendor | Anysphere Inc (San Francisco, USA) |
| Country / origin | 🇺🇸 USA |
| Recommended for Australian users? | ✅ Yes — fully accessible from AUS, AUD card accepted, no region locks |
| Privacy summary | Free tier: Privacy Mode OFF by default; opt in. Pro: Privacy Mode opt-in. Business: Privacy Mode ON by default + no training. |
| Free tier | Yes — limited (Hobby plan); free students get Pro for free for verified students |
| Paid tiers | Pro US40/seat/mo (annual) or US$45 monthly · Enterprise quoted |
| First released | 2023 (initial Cursor IDE); Anysphere founded 2022 |
| Last reviewed | 2026-06-26 |
| Official site | https://cursor.com |
What it is
Cursor is an AI-first code editor. Technically, it’s a fork of Microsoft’s open-source VS Code (Visual Studio Code) editor, but rebuilt with AI deeply integrated into every part of the editing experience — not as a sidebar plugin, but as the core interaction model.
Anysphere — the company behind Cursor — was founded in 2022 by Michael Truell, Sualeh Asif, Arvid Lunnemark, and Aman Sanger (all MIT grads). It’s grown rapidly; by 2026 it’s one of the most-recognised AI coding tools, second only to GitHub Copilot in market share among IDEs.
Cursor’s core differentiators:
- Tab autocomplete that predicts multi-line edits, not just the next character — its proprietary completion model is widely considered the best in class
- Cmd-K (Mac) / Ctrl-K (Windows) in-line AI editing — highlight code, press shortcut, describe the change, get an inline edit
- Cmd-L / Ctrl-L — chat sidebar with codebase context
- Agent mode (Cursor Composer) — multi-file, multi-step AI agent inside the editor (the Cursor answer to Claude Code’s terminal agent)
- Pick-your-model — Claude (Sonnet, Opus, Haiku), GPT-5, GPT-5 Pro, Gemini, Llama, Mistral, plus Cursor’s own optimised completion model
- MCP support — connects to Anthropic’s Model Context Protocol servers
- Privacy Mode — for organisations that need code-never-stored-server-side
Analogy: if VS Code is “a really good text editor with plugins,” Cursor is “VS Code where AI is the primary collaborator and the editor is built around making AI collaboration smooth.” It’s the editor of choice for developers who do most of their coding with AI as a real-time pair, rather than as an occasional helper.
What you’d use it for
- Day-to-day coding with AI as primary collaborator
- Tab autocomplete for fluid multi-line edits as you type
- Inline AI rewriting — “make this function async,” “convert this to TypeScript,” “extract this into a class”
- Multi-file AI agent tasks via Cursor Composer — “rename this component everywhere and update all imports”
- Codebase Q&A — chat with full codebase context
- Test generation — “write tests for this function”
- Documentation generation — “add JSDoc / docstrings to this file”
- Refactoring — large-scale across files
- Working with multiple model providers in one IDE — switch from Claude Sonnet to GPT-5 to Gemini per task
How to sign up + first 5 minutes from Australia
- Go to
cursor.com. Download the Cursor app for Mac, Windows, or Linux. - Install and open Cursor — the UI looks like VS Code because it is a fork.
- Sign in with email / Google / GitHub.
- If you already use VS Code, you can import all your VS Code settings and extensions when prompted on first launch — Cursor is compatible with most VS Code extensions.
- Free tier (Hobby) active immediately: 2,000 completions, 50 slow premium model requests per month.
- Try one of these:
- Open any code file. Start typing. Press Tab to accept Cursor’s autocomplete suggestions (try a function signature — Cursor will often complete the body, not just the line).
- Highlight some code. Press Cmd-K (Mac) or Ctrl-K (Windows). Type “convert this to use async/await.” Watch Cursor rewrite it inline.
- Press Cmd-L (Mac) or Ctrl-L (Windows) to open the chat sidebar. Try: “explain what this file does.”
- Open Cursor Composer (Cmd-I / Ctrl-I) and ask for a multi-file change.
- Optional — upgrade to Pro. Settings → Account → Upgrade → Pro. US$20/mo, AUD card accepted.
- For students: apply for free Pro at cursor.com/students with your
.edu.auemail — verified students get Pro free.
What it costs — what you actually get
Hobby (free)
- 2,000 tab completions / month
- 50 slow premium model requests / month
- Codebase Q&A
- Limited Composer (Agent mode) usage
- Privacy Mode opt-in
Pro — US36 incl GST)
- Unlimited fast tab completions
- 500 fast premium model requests / month (then slow)
- Unlimited slow premium model requests
- Full Composer / Agent mode
- All model providers (Claude, GPT, Gemini, etc.)
- Privacy Mode opt-in
- Higher MCP limits
Business — US45/seat/month (monthly)
- Everything in Pro
- Privacy Mode ON by default
- No training on your code
- SAML SSO
- Admin console
- Centralised billing
- Higher rate limits during peak hours
- Audit logs
Enterprise — quoted
- Everything in Business
- SOC 2 audit reports
- Customer-managed encryption keys
- Custom data residency
- Priority support
- Volume pricing
Free for students
- Apply at
cursor.com/studentswith.edu.auemail - Get Pro free while a verified student
- Renewable annually
Hidden costs to know about
- “Slow” vs “fast” premium model requests — fast uses paid model API time at higher quality / lower latency; slow uses cheaper model or queue. Practical impact: heavy users want Pro.
- Cursor’s own “Cursor Small” completion model is unlimited on Pro and provides the bulk of completion experience; premium-model requests are for Cmd-K / Composer / Chat that use bigger models.
- The Composer agent can rack up many premium-model requests per task — watch usage if cost matters.
How it compares to alternatives
| Capability | Cursor | Claude Code | Windsurf | GitHub Copilot |
|---|---|---|---|---|
| Surface | Full IDE (VS Code fork) | Terminal CLI | Full IDE (VS Code fork) | Plugin in VS Code, JetBrains, Vim, etc. |
| Tab autocomplete | Best (proprietary model) | N/A (terminal) | Excellent | Excellent (Copilot completion) |
| Inline AI edit | Cmd-K (excellent) | N/A (terminal) | Excellent | Inline Copilot Chat |
| Multi-file agent | Composer (excellent) | Best (Claude Code agent) | Cascade (excellent) | Copilot agent mode |
| Codebase awareness | Excellent | Excellent (full repo context) | Excellent | Improving |
| Model picker | Claude, GPT, Gemini, Llama, Mistral, custom | Claude family (deepest integration) | Multiple models | Multiple models (newer) |
| MCP support | Yes | Yes (built by Anthropic) | Yes | Yes (newer) |
| Terminal integration | Yes (built-in terminal) | Native | Yes (built-in terminal) | Yes |
| Cost (consumer) | US$20 | Included in Claude Pro/Max | US$15 | US$10 |
| Best for | Editor-first AI workflow | Terminal-first AI workflow | Editor-first alternative | Lighter plugin-style help |
The honest take: Cursor and Claude Code are complementary, not competitors. Many developers use both — Claude Code in the terminal for agentic / long-running / multi-file tasks; Cursor in the editor for fluid in-line editing and tab completion. If forced to pick one, the choice depends on whether you’re an editor-first or terminal-first developer.
GitHub Copilot is the lighter alternative — better integration with existing VS Code / JetBrains setup, less aggressive editor restructuring than Cursor.
Windsurf is a direct Cursor competitor with very similar capabilities.
See cursor-vs-claude-code-vs-windsurf.md 🟥 for the deeper comparison.
Privacy / data handling
Free tier (Hobby): Privacy Mode is OFF by default. This means Cursor may retain your code, prompts, and AI responses; may use them to improve the product; may share parts with model providers under their standard terms (which include training-by-default for some consumer-tier API calls).
To enable Privacy Mode: Cursor Settings → General → Privacy Mode → ON.
When Privacy Mode is ON:
- Cursor does not retain your code beyond the time needed to serve a request
- Model providers receive code only for the active request and do not store / train (per Cursor’s agreements with them)
Pro: Privacy Mode is opt-in (default OFF unless you enable). Same behaviour as Hobby once enabled.
Business / Enterprise: Privacy Mode is ON by default. Cursor contractually does not train on Business code, does not share beyond the active request, tenant-isolated.
MCP servers: when you use MCP, your inputs / outputs flow through the configured MCP server — verify each MCP server’s own privacy posture.
Where data lives: US data centres primarily; Enterprise can negotiate data-residency.
For sensitive code, business code, or anything client-confidential: turn on Privacy Mode OR use the Business tier. Don’t paste secrets into chat regardless.
Recent changes (since the last review)
- 2026: Claude Sonnet 4.6 / Opus 4.8 + GPT-5 / GPT-5 Pro + Gemini 3 Pro / Flash all available in model picker
- 2026: MCP support expanded; Cursor can connect to most MCP servers
- 2025: Cursor Composer (Agent mode) launched and matured
- 2025: Business tier launched
- 2025: Cursor reached US$10B+ valuation; rapid hiring
- 2024: Tab completion model significantly improved
- 2024: Cursor reached >1M daily active users
- 2023: Initial Cursor launch (fork of VS Code with AI integration)
(Check cursor.com/changelog for the latest.)
Gotchas
- Cursor is a SEPARATE app from VS Code — installing Cursor doesn’t replace VS Code; they coexist. You can use both, but you’ll be running two editors.
- VS Code settings import works on first launch — but if you skip that, retrofitting later requires manual config or extension transfer.
- Most VS Code extensions work in Cursor, but a few don’t (Microsoft GitHub Copilot extension is intentionally blocked since Cursor is a Copilot alternative; some other Microsoft-proprietary extensions may not work).
- Tab autocomplete is aggressive by default — for developers who prefer typing without prediction interruptions, the speed can be too much. Turn down via Settings → Cursor Tab.
- Cmd-K / Ctrl-K can rewrite more than you intended if your selection is large. Use small selections for surgical edits.
- Composer agent can change many files at once — always review the diff before accepting. Don’t run Composer on production code without supervision.
- The “Slow” requests on free / Pro queue when the system is busy — for time-sensitive coding, Business removes the queue.
- Cursor’s pricing model has shifted several times — the encyclopedia notes prices as of last review, but verify on cursor.com.
- Anysphere has had press cycles about model selection — some model providers (notably OpenAI) have at various times pulled or restricted access to Cursor. Pricing and model availability can shift.
- Cursor periodically retrains its proprietary completion model — quality may drift up or down between updates.
- Cursor’s MCP marketplace is less mature than Anthropic’s — for the broadest MCP server ecosystem, Claude Code is still ahead.
- macOS users: Cursor sometimes uses significant memory; check Activity Monitor if your Mac feels sluggish.
See also
- Claude Code deep dive 🟩 🟦 — terminal-first alternative
- Windsurf 🟥 — direct Cursor competitor
- GitHub Copilot 🟥 — lighter VS Code plugin alternative
- OpenAI Codex (2026 agent) 🟥
- Google Antigravity 🟥
- Cline 🟥 — open-source VS Code agent
- Aider 🟥 — open-source CLI agent
- Zed AI 🟥 — alternative AI-first editor
- Continue.dev 🟥
- Replit Agent 🟥
- Devin (Cognition Labs) 🟥
- v0 by Vercel 🟥
- Bolt.new 🟥
- Lovable 🟥
- Claude Skills 🟥
- Claude Cowork 🟥
- MCP — Model Context Protocol 🟩 🟦
- Decision frameworks — Cursor vs Claude Code vs Windsurf 🟥
- which-ai-for-which-job.md 🟩 🟦
- Glossary — C (Cursor) 🟩