Cursor
Made by: Anysphere Type: AI-First IDE (Coding Assistant) Best for: Active development sessions, real-time pair programming, teams moving fast
What it is
Cursor is a code editor built from the ground up for working with AI. It looks and feels like VS Code (it's a fork) but every feature — autocomplete, chat, code generation, refactoring — is designed around AI models rather than bolted on as an extension.
If VS Code with Copilot is a car with a GPS added on, Cursor is a car designed around the GPS from day one.
The core features
Tab completion (Cursor Tab)
As you type, Cursor predicts not just the next token but the next logical edit — often a multi-line change across several locations. Press Tab to accept, Escape to reject. It learns your patterns over time.
Inline chat (Cmd+K)
Select any code and press Cmd+K (Mac) or Ctrl+K (Windows/Linux). Describe what you want:
- "Rewrite this to use async/await"
- "Add error handling"
- "Extract this into a separate function"
Cursor applies the change inline, with a diff you can accept or reject.
Chat panel (Cmd+L)
A sidebar chat that has full context of your codebase. Ask it questions:
- "How does authentication work in this app?"
- "Where is the user's email stored?"
- "Explain what this function does"
Or give it tasks that span multiple files. Unlike a plain chatbot, Cursor's chat can directly apply changes to your code.
Composer (Cmd+I)
The most powerful mode — a full-screen agent that can create and edit multiple files at once. Describe a feature end-to-end and Composer plans and executes it.
Codebase context: @-mentions
Cursor lets you pull specific context into any conversation using @:
| Mention | What it includes |
|---|---|
@Files | A specific file |
@Folders | An entire directory |
@Code | A specific function or class |
@Docs | External documentation (React, Next.js, etc.) |
@Web | A web search result |
@Git | Commit history or a diff |
@auth.ts Why is the token being rejected on refresh?Which model runs Cursor?
Cursor works with multiple models — you choose in settings:
- Claude Sonnet / Opus — best for complex reasoning and large context
- GPT-4o — fast, good at code
- o1 / o3 — deep reasoning for hard algorithmic problems
- Cursor's own models — fine-tuned for autocomplete speed
Most teams default to Claude Sonnet for chat and keep Cursor's fast model for Tab autocomplete.
Cursor vs Claude Code
| Cursor | Claude Code | |
|---|---|---|
| When to use | Active coding session | Delegated tasks, CI/CD |
| Interface | GUI IDE | Terminal |
| Collaboration feel | Pair programmer | Junior engineer you delegate to |
| Speed of feedback | Instant, inline | End-to-end (plan → build → test) |
| Best for | Frontend work, exploratory development | Backend refactors, automation |
Many developers use both — Cursor for the active coding session, Claude Code for the grunt work they'd rather not do themselves.
Pricing
- Free tier — 2,000 completions/month + 50 slow premium requests
- Pro ($20/month) — Unlimited fast completions + 500 premium requests
- Business ($40/user/month) — Team features, SSO, admin controls, privacy mode
Privacy mode disables training on your code. Recommended for proprietary codebases. Available on Pro and Business plans.
Getting started
- Download from cursor.sh (opens in a new tab)
- It imports your VS Code settings, extensions, and keybindings automatically
- Sign in and pick your preferred model
- Open a project and start typing — Cursor Tab is on by default
The fastest way to feel the difference: open a file you haven't touched in a while and ask @file What does this do and what could go wrong?