Quick Reference

Reference intermediate

Cursor Quick Reference Card

Three AI Modes

Mode Shortcut Best For Scope
Inline Edit Cmd+K / Ctrl+K Quick targeted changes Selected code
Chat Cmd+L / Ctrl+L Questions, single-file generation Conversation
Agent Cmd+I / Ctrl+I Multi-file features, agent tasks Project-wide

Essential Shortcuts (Mac / Windows)

Action Shortcut
AI Chat Cmd+L / Ctrl+L
Inline Edit Cmd+K / Ctrl+K
Agent Cmd+I / Ctrl+I
Accept suggestion Tab
Reject suggestion Esc
Command Palette Cmd+Shift+P / Ctrl+Shift+P
Quick Open Cmd+P / Ctrl+P
Toggle Terminal Ctrl+`

Context References (@-mentions)

Reference What It Does When to Use
@file Includes specific file Need exact content
@folder Includes folder structure + contents Need broad awareness
Agent's semantic search Semantic search across project Finding patterns
@docs Searches indexed documentation Need API details
`` Live web search Need current info
@git Git history and diffs Need change context
`` Symbol definitions Need type info

Project Configuration Files

File Location Purpose
.cursor/rules Project root AI behavior rules
.cursorignore Project root Exclude from AI context
.cursor/rules/*.mdc .cursor/rules/ Directory-specific rules

Workflow Quick Guides

Quick Fix:

Select code > Cmd+K > "Fix [issue]" > Tab

Generate Component:

Cmd+L > "@file:types.ts Create a component that..." > Review > Accept

Multi-file Feature:

Cmd+I > Describe feature + @file references > Review diffs > Accept All

Understand Code:

Select code > Cmd+L > "Explain this code and its edge cases"

Model Selection Guide

Task Recommended Model Why
Quick edits Fast model (e.g., cursor-small) Speed matters
Complex logic A frontier reasoning model (e.g. Claude Opus/Sonnet) Reasoning quality
Architecture Claude Opus Deep analysis
Boilerplate Any model All handle well

.cursor/rules Minimal Template

# Project: [Name]
# Stack: [Language] + [Framework] + [Key Libraries]

## Conventions
- [Convention 1]
- [Convention 2]
- [Convention 3]

## Patterns
- [Pattern to follow]

## Do NOT
- [Anti-pattern 1]
- [Anti-pattern 2]

Accept/Reject Decision Flow

AI generates code
       │
  Does it compile? ──No──> Reject, provide error
       │
      Yes
       │
  Correct logic? ──No──> Refine: "Fix [specific issue]"
       │
      Yes
       │
  Matches style? ──No──> Refine: "Match project style"
       │
      Yes
       │
  Accept (Tab / Accept All)

Tips for Better Results

  1. Open related files in tabs before prompting (Cursor uses them as context)
  2. Be specific about what you want changed vs. left alone
  3. Use @file for exact context rather than relying on Agent's semantic search
  4. Start new chats when switching topics (avoids context pollution)
  5. Check the model you are using matches the task complexity
  6. Save files before referencing them in chat
  7. Use Agent for anything touching 3+ files