Cursor

# Open AI Chat
Ctrl+I         # Toggle Chat panel
Ctrl+L         # Add selection to new chat

# Chat controls
Enter          # Send message (Nudge mode)
Ctrl+Enter     # Queue/Force send message
Ctrl+Shift+BackSpace  # Cancel generation

# Navigation
Ctrl+N         # New chat
Ctrl+T         # New chat tab
Ctrl+[         # Previous chat
Ctrl+]         # Next chat
Ctrl+W         # Close chat

// EXAMPLE: Working with AI Chat
1. Select problematic code
2. Press Ctrl+L to add to chat
3. Type: "Fix the type errors in this function"
4. Press Enter to send
5. Ctrl+Enter to accept all changes

# Inline editing
Ctrl+K         # Open inline edit input
Ctrl+Shift+K   # Toggle input focus/Add selection

# Quick actions
Alt+Enter      # Ask quick question about code
Ctrl+Shift+Enter  # Perform full file edits

# Controls
Enter          # Submit edit request
Ctrl+Shift+BackSpace  # Cancel edit
Tab            # Jump to next edit location

// EXAMPLE: Refactoring with Inline Edit
1. Select function to refactor
2. Press Ctrl+K
3. Type: "Convert to async/await"
4. Press Enter
5. Review changes inline
6. Tab to jump between edits

# Accept suggestions
Tab            # Accept full suggestion
Ctrl+→         # Accept word-by-word
Esc            # Reject suggestion

# After edits
Tab            # Jump to next predicted location

# Configuration
Cursor Settings > Features > Tab
- Enable/disable in comments
- Configure auto-import
- Set file extension exclusions

// EXAMPLE: Smart Tab Completion
// Start typing a function...
function calc
// Tab accepts: function calculateTotal(items) {
//   return items.reduce((sum, item) => sum + item.price, 0)
// }
// Tab again jumps to next likely edit location

# Selection context
Ctrl+Shift+L   # Add selected code to chat
Ctrl+L         # Add selection to new chat

# Clipboard context
Ctrl+V         # Add clipboard as context (auto-detect)
Ctrl+Shift+V   # Add clipboard to input box

# Reference operators
@              # Reference specific items
#              # Reference files
/              # Use shortcut commands

# Codebase search
Ctrl+Enter     # Search entire codebase in chat

// EXAMPLE: Multi-file context
> @file1.ts @file2.ts How do these interact?

// EXAMPLE: Codebase search
> Find all API endpoints [Ctrl+Enter]

// EXAMPLE: Reference specific function
> @calculateTotal Why does this return NaN?

# Rule file locations
.cursor/rules/         # Project-specific rules
AGENTS.md             # Simple project rules
User Settings > Rules  # Global user rules

# Rule file format (.cursor/rules/*.mdc)
---
description: API Standards
globs: ["**/*.api.ts"]
alwaysApply: false
---
- Use REST conventions
- Include error handling
- Add JSDoc comments

# Legacy: root .cursorrules still loads but is
# ignored in Agent mode - migrate to .cursor/rules/*.mdc

# Rule behaviors
Always         # Always included in context
Auto Attached  # Included with matching files
Agent Requested # AI decides inclusion
Manual         # Only when mentioned

// EXAMPLE: Project rules in .cursor/rules/style.md
---
description: Code style guide
alwaysApply: true
---
- Use TypeScript strict mode
- Prefer const over let
- Use async/await over callbacks
- Add error boundaries to components

# Memories are automatically generated
- Created from chat conversations
- Project-scoped context
- User approval required

# Management
Cursor Settings > Rules > Memories
- View all memories
- Edit or delete memories
- Control auto-generation

# Memory format
"Always use error-first callbacks in Node.js"
"Prefer named exports over default exports"
"User prefers functional components"

// EXAMPLE: Memory generation
You: "Always use Zod for validation"
Cursor: "I'll remember to use Zod for validation"
[Memory saved for project]

// Future conversations:
You: "Add validation to this API"
Cursor: [Automatically uses Zod based on memory]

# Quick model switching
Ctrl+/         # Loop between models
Ctrl+Alt+/     # Model toggle in chat

# Available providers
- Cursor models (default)
- OpenAI (GPT-4, GPT-3.5)
- Anthropic (Claude 3.5, Claude 3)
- Google (Gemini)
- DeepSeek
- xAI
- Azure OpenAI
- AWS Bedrock

# Special modes
Auto Mode      # Automatically selects best model
Max Mode       # Extended context window
Privacy Mode   # No data storage

// EXAMPLE: Switching models for different tasks
Ctrl+/ → GPT-4 for complex logic
Ctrl+/ → Claude for refactoring
Ctrl+/ → Cursor-small for quick edits

// EXAMPLE: Auto mode selection
Auto Mode → Uses o1 for complex algorithms
Auto Mode → Uses GPT-4 for general coding

# Configure in Cursor Settings > Models

# Supported custom keys:
- OpenAI API key
- Anthropic API key
- Google AI API key
- Azure OpenAI endpoint
- AWS Bedrock credentials

# Configuration steps:
1. Open Cursor Settings (Ctrl+Shift+J)
2. Navigate to Models
3. Enter API key
4. Verify connection

# Notes:
- Keys encrypted, not stored locally
- Custom keys don't support o1/o1-mini
- Usage billed to your account

// EXAMPLE: Azure OpenAI setup
{
  "azure": {
    "endpoint": "https://myorg.openai.azure.com",
    "apiKey": "your-api-key",
    "deployment": "gpt-4"
  }
}

// EXAMPLE: Using custom OpenAI key
Settings > Models > OpenAI
API Key: sk-...
✓ Verified and connected

# Access Agent Mode
Ctrl+I         # Open chat
Select "Agent" mode from dropdown

# Agent capabilities:
- Semantic codebase search
- Multi-file editing
- Terminal command execution
- Automatic change tracking
- Creates checkpoints

# Best for:
- Feature implementation
- Bug fixes across files
- Refactoring projects
- Setting up configurations

// EXAMPLE: Agent implementing feature
You: "Add user authentication with JWT"

Agent:
1. Searching for existing auth code...
2. Creating auth middleware...
3. Setting up JWT utilities...
4. Adding login/logout endpoints...
5. Updating user model...
6. Running tests...
✓ Authentication system implemented

# Enable Background Agent
Ctrl+E         # Open Background Agent panel

# Features:
- Creates remote VM
- Explores project automatically
- Detects bugs proactively
- Proposes fixes
- Runs independently

# Configuration:
Settings > Features > Background Agent
- Enable/disable
- Set inspection frequency
- Configure notification preferences

// EXAMPLE: Background Agent findings
[Background Agent Notification]
"Found potential null reference in auth.service.ts:45"
→ View Issue
→ Apply Fix
→ Dismiss

// Automatically discovers:
- Type errors
- Security vulnerabilities
- Performance issues
- Code smells

# Open command palette
Ctrl+Shift+P   # Command palette

# Useful commands:
"Cursor: Settings"      # Open Cursor settings
"Cursor: Export Chat"   # Export to markdown
"Cursor: Clear Cache"   # Clear AI cache
"Cursor: Toggle Privacy Mode"
"Cursor: Show Rules"
"Cursor: Snooze Tab"    # Temporarily disable

# Custom commands:
Add to keybindings.json for shortcuts

// EXAMPLE: Common command workflows
Ctrl+Shift+P → "Export Chat"
→ Saves conversation as markdown

Ctrl+Shift+P → "Toggle Privacy"
→ Enables privacy mode for sensitive code

Ctrl+Shift+P → "Clear Cache"
→ Fixes stuck suggestions

# Available modes (dropdown in chat):

Agent Mode     # Autonomous task completion
Ask Mode       # Question-answering only
Custom Modes   # User-defined behaviors

# Creating custom modes:
Settings > Features > Custom Modes
{
  "name": "Code Review",
  "prompt": "Review code for bugs, security issues, and improvements",
  "temperature": 0.3
}

# Mode shortcuts:
Ctrl+.         # Quick mode menu

// EXAMPLE: Custom Security Mode
{
  "name": "Security Audit",
  "prompt": "Check for vulnerabilities, validate inputs, ensure encryption",
  "includeFiles": ["*.auth.*", "*.api.*"],
  "temperature": 0.1
}

// Usage:
Select "Security Audit" mode
> "Review this endpoint"

# Available integrations:

GitHub         # PR reviews, issue tracking
Git            # Version control integration  
Slack          # Team notifications
Linear         # Task management
Bugbot         # Automated code review

# Setup:
Settings > Integrations > [Service]
- Authenticate
- Configure permissions
- Set notification preferences

# Shell commands:
cursor://command/<command>
- Custom URL protocol
- Trigger from external tools

// EXAMPLE: GitHub integration
1. Connect GitHub account
2. Enable PR reviews
3. Bugbot automatically reviews PRs
4. Comments appear in GitHub

// EXAMPLE: Slack notifications
"Bug found in payment.service.ts"
→ Posted to #dev-alerts channel

# Best practices:

1. Use specific file references
   @auth.service.ts not "the auth file"

2. Include error messages
   Ctrl+V to paste errors as context

3. Create project rules
   .cursor/rules/ for consistency

4. Review memories periodically
   Settings > Rules > Memories

5. Clear context when switching tasks
   Start new chat for different features

// GOOD: Specific context
> @user.model.ts @auth.service.ts 
> "Add password reset with email verification"

// BAD: Vague context
> "Fix the login"

// GOOD: Include errors
> [Paste TypeError stack trace]
> "Fix this type error"

# Performance tips:

1. Disable Tab for large files
   Settings > Features > Tab > Exclusions
   Add: "*.min.js", "*.bundle.js"

2. Use Privacy Mode for sensitive code
   Prevents indexing of confidential files

3. Clear cache if suggestions lag
   Ctrl+Shift+P → "Clear Cache"

4. Snooze Tab when not needed
   Temporarily disable completions

5. Limit context in large projects
   Use specific @ references vs entire folders

// EXAMPLE: Exclusion patterns
{
  "tab.exclusions": [
    "*.min.js",
    "*.bundle.js", 
    "node_modules/**",
    "dist/**",
    "*.generated.ts"
  ]
}

# Common issues and solutions:

"Suggestions not appearing"
→ Check Tab is enabled
→ Clear cache (Ctrl+Shift+P)

"Chat not responding"
→ Check model selection
→ Verify API key if custom

"Wrong context included"
→ Review @ references
→ Check .cursor/rules/

"Agent stuck"
→ Ctrl+Shift+BackSpace to cancel
→ Review checkpoint history

"High memory usage"
→ Disable Background Agent
→ Limit open chat tabs

// EXAMPLE: Debug process
1. Ctrl+Shift+P → "Developer: Toggle Developer Tools"
2. Check Console for errors
3. Network tab for API issues
4. Report via Settings > Help > Report Issue