Bookmark this page. Use Ctrl+F (or Cmd+F on Mac) to find what you need. This cheat sheet compares AI coding tools, their features, pricing, and best use cases.

Last updated: March 2026

Tool Comparison

ToolTypeBest ForPricing (Mar 2026)
Claude CodeCLI agentMulti-file edits, refactoring, debugging$20/mo Pro, $100/$200 Max
CursorIDE (VS Code fork)Day-to-day coding with AI inline$20/mo (Pro)
GitHub CopilotIDE extensionAutocomplete, inline suggestionsFree (limited), $10/mo Pro, $39/mo Pro+
WindsurfIDE (VS Code fork)Chat + agent workflows$15/mo (Pro)
ChatGPTChatExplaining code, brainstorming$20/mo (Plus)
Gemini CLICLIQuick questions, code reviewFree (1K req/day), Pro/Ultra plans
Codex CLICLICode review, autonomous tasksOpenAI API pricing

Feature Matrix

FeatureClaude CodeCursorCopilotWindsurf
AutocompleteYesYesYes
ChatYesYesYesYes
Multi-file editingYesYesLimitedYes
Terminal commandsYesYes (Cmd+K in terminal)
Agent modeYesYesYesYes
Git integrationYesYesYesYes
MCP supportYesYesYesYes
Custom rulesCLAUDE.md.cursorrules.github/copilot-instructions.md.windsurfrules
Context window200K (up to 1M on Max)128K128K128K

Claude Code Shortcuts

CommandDescription
/helpShow available commands
/clearClear conversation
/compactCompress context to save tokens
/costShow token usage and cost
/initCreate CLAUDE.md project file
EscCancel current generation
Shift+TabAccept file edit
Ctrl+CExit

Claude Code Tips

# Run Claude Code
claude

# Start with a prompt
claude "fix the failing tests"

# Non-interactive mode
claude -p "explain this function" < file.py

# Use a specific model
claude --model sonnet   # or opus, haiku

# Continue last conversation
claude --continue

CLAUDE.md — Project Instructions

# CLAUDE.md
- Use TypeScript strict mode
- Tests: Jest with React Testing Library
- Style: Tailwind CSS, no inline styles
- Always run tests before committing

Place CLAUDE.md in your project root. Claude reads it automatically.

Cursor Shortcuts

ShortcutAction
Cmd+K / Ctrl+KInline edit (AI rewrites selected code)
Cmd+L / Ctrl+LOpen AI chat panel
Cmd+I / Ctrl+IComposer (multi-file agent)
TabAccept autocomplete suggestion
EscDismiss suggestion
Cmd+Shift+PCommand palette
@fileReference a file in chat
@codebaseSearch entire codebase for context
@webSearch the web
@docsReference documentation

.cursorrules — Project Instructions

You are a senior TypeScript developer.
- Use functional components with hooks
- Prefer Tailwind CSS utility classes
- Write tests for every new function

GitHub Copilot Shortcuts

ShortcutAction
TabAccept suggestion
EscDismiss suggestion
Alt+] / Option+]Next suggestion
Alt+[ / Option+[Previous suggestion
Ctrl+EnterOpen Copilot completions panel
Cmd+I / Ctrl+IInline chat (Copilot Chat)

Effective Prompting Tips

TipExample
Be specific“Add error handling to the fetchUsers function in api.ts” not “fix the code”
Give context“This is a Next.js 14 app using App Router and Prisma ORM”
Specify output“Return only the modified function, not the full file”
Reference files“Look at src/types.ts for the User interface”
Ask for alternatives“Show me 3 different ways to implement this”
Iterate“Good, but use async/await instead of .then()”

When to Use Which Tool

ScenarioBest Tool
Quick inline autocomplete while typingCopilot or Cursor
Large refactoring across many filesClaude Code
Explaining unfamiliar codeChatGPT or Claude Code
Building a new feature from scratchCursor Composer or Claude Code
Code reviewCodex CLI or Claude Code
Debugging a tricky bugClaude Code (terminal access)
Writing documentationChatGPT or Claude Code
Learning a new frameworkChatGPT (conversational)

MCP (Model Context Protocol)

MCP lets AI tools connect to external services (databases, APIs, Figma):

// .mcp.json in project root
{
  "mcpServers": {
    "postgres": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-postgres"],
      "env": { "DATABASE_URL": "postgresql://..." }
    }
  }
}

Supported by: Claude Code, Cursor, Copilot, Windsurf. Lets the AI query your database, read Figma designs, or interact with any MCP-compatible service. See our MCP Protocol Explained article for details.

Common Mistakes

  1. Accepting code without reviewing — AI suggestions can contain bugs, security issues, or outdated patterns. Always read the diff before accepting. Treat AI output as a junior developer’s first draft.

  2. Not providing enough context — “Fix the bug” gives bad results. “Fix the null pointer in handleSubmit at line 42 of LoginForm.tsx — the user object can be undefined when the session expires” gives great results.

  3. Using AI for everything — Simple tasks (renaming a variable, adding an import) are faster to do manually. Use AI for complex tasks where it saves significant time.