Claude Code Cheatsheet

Find the right command, file, or shortcut — fast.

Last verified: 2026-04-06

Start Fast

# macOS / Linux / WSL (official native installer)
curl -fsSL https://claude.ai/install.sh | bash

# macOS alternative
brew install --cask claude-code

# Windows
winget install Anthropic.ClaudeCode

Authenticate and verify the setup:

claude auth login
claude auth status --text
claude                   # interactive session
claude "explain this repo"  # start with a prompt
claude -c                # continue last conversation
claude --resume          # pick a session to resume

Type /help inside any session to see all available commands.

This course is terminal-first, but Claude Code also runs in VS Code, JetBrains, Desktop, and Web.

Daily Commands

Command	Description
`claude`	Start an interactive session
`claude -p "prompt"`	One-shot print mode (non-interactive)
`claude -c`	Continue the last conversation
`/resume`	Re-open a saved conversation
`/clear`	Reset the conversation
`/compact`	Compress context to free up tokens
`/cost`	Check token usage and spend
`/status`	See version, model, account, and connectivity
`/diff`	Review current and per-turn diffs
`/permissions`	Manage tool access for the session

CLI Commands

Command	Description
`claude auth login`	Sign in
`claude auth status --text`	Check authentication state
`claude update`	Update Claude Code
`claude agents`	List configured sub-agents
`claude mcp`	Manage MCP servers from the CLI
`claude plugin`	Manage plugins
`claude remote-control --name "my session"`	Start a Remote Control server session

Slash Commands

Daily

Command	Description
`/help`	Show all commands
`/resume`	Resume a session
`/rename "name"`	Rename this session
`/clear`	Start fresh
`/compact`	Compress context
`/cost`	Check usage
`/status`	Show version, model, account, and connectivity
`/permissions`	Manage tool access

Advanced

Command	Description
`/plan`	Enter plan mode (read-only)
`/init`	Generate CLAUDE.md for this project
`/memory`	View/edit loaded instructions and auto-memory
`/model`	Inspect or switch models
`/effort`	Adjust reasoning depth
`/context`	Visualize context usage
`/diff`	Open interactive diff viewer
`/doctor`	Diagnose install and settings issues
`/hooks`	Inspect active hook configuration
`/mcp`	Manage MCP connections and auth
`/tasks`	Open the task list for agent teams and ultraplan sessions
`/skills`	List available skills
`/plugin`	Manage plugins
`/agents`	List configured sub-agents
`/add-dir`	Add directory access
`/config`	Configure settings
`/remote-control`	Continue this session remotely
`/ultraplan "prompt"`	Send a planning task to Claude Code on the web for richer review
`/security-review`	Review current diff for security issues

Automation

Command	Description
`/loop 5m "prompt"`	Repeat a prompt on an interval
`/schedule`	Create cloud scheduled tasks
`/batch "instruction"`	Parallel changes across files
`/install-github-app`	Set up the GitHub Actions integration

Claude Code shows built-in commands and bundled skills together in the slash menu. High-value bundled skills include /simplify, /batch, /debug, and /loop.

/ultraplan is in research preview. It requires Claude Code on the web and a GitHub repository. Launch it with /ultraplan <prompt>, include the word ultraplan in a normal prompt, or escalate a local plan to the web from the plan approval dialog. Use /tasks to open the active cloud session, comment inline in the browser, then either execute on the web or send the approved plan back to your terminal.

Experimental

Agent teams – coordinate multiple sub-agents on a single task. Enable with CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1.

Models & Flags

Common flags

Flag	Description
`--model opus\\|sonnet\\|haiku`	Choose a model
`--effort low\\|medium\\|high\\|max`	Reasoning depth
`-p`	Print mode (non-interactive)
`-c`	Continue last conversation
`-r "<session>"`	Resume a named or ID-based session
`-w [name]`	Work in a git worktree
`-n "name"`	Name the session
`--remote-control`	Start an interactive session with Remote Control enabled

Advanced flags

Flag	Description
`--output-format text\\|json\\|stream-json`	Choose scripting output format
`--max-turns N`	Limit agentic iterations
`--max-budget-usd N`	Cap spend in USD
`--permission-mode`	`default`, `acceptEdits`, `plan`, `auto`, `dontAsk`, `bypassPermissions`
`--add-dir ../path`	Add extra directory to context
`--mcp-config file.json`	Load custom MCP config
`--tools "Bash,Edit,Read"`	Restrict built-in tools for a session
`--bare`	Minimal scripted mode with less auto-discovery

Files & Locations

File	Scope	Purpose
`CLAUDE.md`	Project (shared)	Project-level instructions
`.claude/CLAUDE.md`	Project (shared)	Alternative project location
`CLAUDE.local.md`	Project (gitignored)	Personal overrides
`~/.claude/CLAUDE.md`	User	Instructions for all projects
`.claude/rules/*.md`	Project (shared)	Path-specific rules
`.claude/settings.json`	Project (shared)	Project settings and hooks
`.claude/settings.local.json`	Project (gitignored)	Local settings
`~/.claude/settings.json`	User	User-wide settings
`.mcp.json`	Project (shared)	MCP server config
`.claude/skills/*/SKILL.md`	Project (shared)	Custom skills
`~/.claude/skills/*/SKILL.md`	User	Personal skills
`.claude/agents/*.md`	Project (shared)	Custom sub-agents
`~/.claude/agents/*.md`	User	Personal sub-agents

MCP Setup

claude mcp add <name> --scope project -- <command> [args...]
claude mcp list
claude mcp remove <name>

Project scope stores config in .mcp.json. User scope stores config in ~/.claude.json.

Hooks Basics

Hooks live in .claude/settings.json under the hooks key.

Hook types: command, prompt, agent, http

Key events

Event	When it fires
`PreToolUse`	Before any tool runs
`PostToolUse`	After a tool succeeds
`SessionStart`	Session begins
`Stop`	Claude finishes responding

Exit codes: 0 = proceed, 2 = block the action.

Hooks receive JSON on stdin. Use jq to extract fields in shell hooks.

GitHub Actions

# .github/workflows/claude.yml
on:
  issue_comment: { types: [created] }
  pull_request_review_comment: { types: [created] }
jobs:
  claude:
    if: contains(github.event.comment.body, '@claude')
    runs-on: ubuntu-latest
    steps:
      - uses: anthropics/claude-code-action@v1
        with:
          anthropic_api_key: $

Advanced

Headless mode – pipe Claude into scripts:

claude -p "list TODO comments" --output-format json

Remote control – continue the same local session from phone or browser:

claude --remote-control "my rollout"

Agent teams – experimental multi-agent coordination:

export CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1
claude "Review security, performance, and quality"

Budget controls:

Flag	Description
`--max-budget-usd N`	Hard spend cap
`--max-turns N`	Limit agentic loops

Keyboard Shortcuts

Key	Action
`Shift+Tab`	Cycle permission modes
`Ctrl+C`	Interrupt current turn
`Ctrl+L`	Clear the terminal screen
`Ctrl+O`	Toggle verbose mode
`Esc` `Esc`	Edit the previous message
`Ctrl+T`	Toggle task list (agent teams)
`Shift+Down`	Cycle teammates
`/`	Open slash command autocomplete