Cursor AI Editor Cheat Sheet: Features, Shortcuts & Workflows

Quick reference tables

Installation & setup

Core AI commands

Composer (Cmd+L → New Tab)

The Composer tab handles complex multi-file changes:

Context references (@)

Attach files, folders, docs, or symbols to any prompt:

# Example: explain a file with its test and recent commits
@file:src/auth/login.ts @file:src/auth/login.test.ts @git:5

Inline edit workflow

1. Select code or place cursor where you want changes
2. Press Cmd+K → type your instruction
3. Cursor shows a diff preview (highlighted in red/green)
4. Press Tab to accept, Esc to reject
5. Press Cmd+Z to undo the AI change

Tab autocomplete settings

Cmd+, → Cursor → Tab → Configure behavior:

Rules for AI (Project Instructions)

Project-level instructions that tell Cursor how to behave:

Cmd+L → settings (top-right gear) → Rules → Add Rule

Or create .cursor/rules/ directory with .md files:

# .cursor/rules/typescript.md

- Always use explicit return types on exported functions
- Prefer `interface` over `type` for object shapes
- Use Zod for runtime validation of external data
- No `any` — use `unknown` + type narrowing

# .cursor/rules/project.md

- Project uses Next.js App Router
- API routes go in `src/app/api/`
- Server Components by default, Client Components use "use client"
- Use Tailwind for all styling
- Environment variables in `.env.local`

Apply rules per folder using .cursor/rules/ or globally via Cmd+, → Rules.

Composer workflows

Write a new feature

/compose

Create a new feature for user notifications:
- Create src/services/notification.ts with send() and schedule() functions
- Add notification API route at src/app/api/notifications/route.ts
- Create a React hook useNotifications in src/hooks/useNotifications.ts
- Add unit tests in src/services/notification.test.ts
- Update src/app/globals.css with notification styles

Refactor a module

/compose

Refactor src/database/queries.ts:
- Convert to use Drizzle ORM patterns
- Add connection pooling
- Move raw SQL to separate migration files
- Keep the same public API (exported functions)
- Preserve all existing tests

Debug a stack trace

Paste the full error stack trace into Composer:

/fix

[Error stack trace here]

The error happens in src/middleware/auth.ts at line 47.
Fix it and explain what caused it.

Debugger integration

Cursor reads your stack traces and generates debugging plans:

1. Run your code until it throws an error
2. Copy the full stack trace from the terminal
3. Open Composer (Cmd+L → New Tab)
4. Paste the stack trace and type /fix
5. Cursor analyzes the trace and suggests:
   • The exact line causing the error
   • What the expected vs actual values are
   • A corrected version of the problematic code
   • An explanation of the root cause

For breakpoints and step-through debugging:

• F5 — Start debugger (standard VS Code)
• F9 — Toggle breakpoint
• F10 — Step over
• F11 — Step into
• Shift+F11 — Step out

Model selection

Cmd+, → Models → Choose default and per-task models:

Composer and Chat can use different models. Tab completions use a separate fast model.

Best practices

Effective prompt writing

Be specific about the outcome, not just the change:

❌ "Fix the login bug"
✅ "When a user submits the login form with an invalid email,
    the error message should appear inline below the email field
    (not in a toast). The field border should turn red (#ef4444).
    Do not change the API route or any server-side logic."

Chain context with @

❌ "@fix the auth bug"
✅ "@file:src/middleware/auth.ts @file:src/app/login/page.tsx
    @file:src/hooks/useAuth.ts
    The auth middleware is rejecting valid tokens on Vercel
    deployment but works locally. Fix it."

Use Rules for consistency

• Add a .cursor/rules/ folder in project root
• One file per convention (typescript.md, styling.md, api.md)
• Rules apply to all AI interactions in that project

Summary

• Cmd+K — Inline AI edit; Tab — Accept completion
• Cmd+L — Chat; Cmd+L → New Tab — Composer
• @ symbols attach context: files, folders, docs, git history
• Rules for AI enforce project conventions in every AI interaction
• Composer handles multi-file features, refactors, and debugging plans
• Tab completions are fast and local; Composer uses cloud models

FAQ

How does Cursor differ from GitHub Copilot? Cursor is a VS Code fork with AI built into every surface — inline edits, chat, composer, Tab completion, and debugger integration all in one. Copilot is an extension running inside a separate editor. Cursor’s Composer can make multi-file changes in one pass; Copilot fills individual completions.

Can I use my own API key? Yes. Cmd+, → Models → Add custom model → enter your API key for OpenAI, Anthropic, or Google.

Does Cursor work with JetBrains IDEs? Cursor is a VS Code fork only. For JetBrains, use the native Junie AI or Codeium.

How do I stop Cursor from suggesting code I don’t want? Use Tab with Esc to reject. For persistent suppression, add the file to .cursorignore or tighten the model in Cmd+, → Models → Tab model.

Can Cursor access the internet? Not natively. Use @docs:url to attach specific URLs as context, or use @search:query to search your codebase only.

Is Cursor free to use? Cursor has a free tier with limited Composer uses per month. Pro is $20/month for unlimited access. Tab completions are always free.

What to read next

• VS Code Shortcuts Cheat Sheet — keyboard shortcuts that work in Cursor too
• Git Cheat Sheet — version control for your AI-edited code
• GitHub Actions Cheat Sheet — CI/CD for projects built with Cursor

Related Articles

Deepen your understanding with these curated continuations.

View All Articles

VS Code5 min read

VS Code Shortcuts for Developers: The Complete Cheatsheet

Master VS Code keyboard shortcuts — editing, navigation, debugging, Git integration, and custom keybindings. Work faster with the editor you already use.

Darsh JariwalaMar 31, 2026

Cheatsheet5 min read

Ollama Cheat Sheet: Local LLMs, Models, API & Integration (2026)

Complete Ollama reference — pull and run local LLMs, API endpoints, Python/JS integration, multimodal models, model management, and GPU setup in 2026.

Darsh JariwalaMay 19, 2026

AI5 min read

Qwen Coder Cheatsheet (2026 Edition): Running Local Agents

Master Alibaba's open-weights Qwen Coder models. Essential commands for Ollama integration, local execution, and private agentic workflows.

VishnuMar 28, 2026