Cursor Cheat Sheet
Last updated: April 2026
Quick Facts
Pricing
Freemium. Free plan with limited AI queries, Pro at $20/month for unlimited, and custom Enterprise pricing.
Free Plan
Yes. Includes 50 GPT-4 queries per month, 200 GPT-3.5 queries, and basic codebase-aware chat. Good for trying it out.
Rating
4.7/5
Best For
Professional developers and teams who want deep, context-aware AI assistance integrated directly into their editor, not just a chat window.
Key Features
- ✓Chat with Your Codebase
I can ask questions about my entire project. Cursor reads all relevant files and provides answers based on the actual code, not just generic examples.
- ✓AI Code Generation (@-Commands)
By typing @ in the chat, I can reference specific files, functions, or directories. This grounds the AI's response in my exact code context.
- ✓Edit in Diff View
When Cursor suggests an edit, it shows a side-by-side diff. I can review, accept, or reject each change line-by-line, which is crucial for control.
- ✓Agent Mode (Cmd/Ctrl+K)
I give a high-level instruction, and Cursor plans and executes multiple steps—editing several files, running terminal commands—to complete the task autonomously.
- ✓Deep Codebase Indexing
Cursor builds a semantic index of my repo. This allows it to answer complex questions about architecture and dependencies that span many files.
- ✓Seamless VS Code Foundation
It feels 95% like VS Code. All my extensions, themes, and keybindings work. The transition was frictionless, which was a huge win for adoption.
- ✓Generate Tests from Code
I can select a function, ask Cursor to write tests, and it generates comprehensive unit tests using my project's preferred testing framework and patterns.
- ✓Explain This Code
Selecting any block and using this command gives me a clear, concise explanation. It's my go-to for understanding legacy or complex code quickly.
- ✓Find and Fix Bugs
I can paste an error message, and Cursor will analyze the stack trace and relevant code to suggest a fix, often pinpointing the exact line.
- ✓Refactor with Simple Instructions
Commands like 'rename this variable globally' or 'extract this logic into a new function' work flawlessly across the entire indexed codebase.
- ✓Integrated Terminal AI
I can ask Cursor for a shell command from within the terminal panel. It suggests the correct command and I can run it with one click.
- ✓Commit Message Generation
When I stage changes, Cursor can analyze the diff and write a detailed, conventional commit message. It saves me a few minutes every time.
Tips & Tricks
Use @-references religiously. Typing @ then a filename forces the AI to ground its answer in that specific code, drastically improving accuracy.
For complex tasks, start in Agent Mode (Cmd+K). Let it draft a plan first, review the steps, then execute. It prevents wild goose chases.
Keep the 'Chat' panel open on the side. I treat it like a senior dev pair programmer, constantly asking small questions as I work.
When it generates code, always use the Diff View. Never blindly accept. This habit catches subtle logic errors and maintains code quality.
Index your entire codebase in Settings. The initial scan takes time but is worth it for the deep, context-aware answers you'll get later.
Common Commands
Cmd/Ctrl + KOpens the Agent command box to give a high-level instruction for a multi-step task.
Cmd/Ctrl + LOpens the Chat panel to ask questions about your code or generate new code.
Cmd/Ctrl + IWith code selected, opens a menu of quick actions like Explain, Refactor, or Generate Tests.
Limitations
- -The free plan's 50 GPT-4 queries run out very fast if you're actively developing. You'll feel the limit within a day or two.
- -Indexing large monorepos can be slow and sometimes unstable. I've had to re-index projects over 5GB, which is a pain.
- -It can still hallucinate library APIs or project specifics. You must have the knowledge to vet its output; it's not infallible.
- -The AI sometimes gets 'stuck' in a suboptimal solution during Agent mode, requiring manual intervention to correct its course.