A practical guide to every CLI command, flag, and in-session slash command you need to get productive with Claude Code fast.
Think of this as the Claude Code command cheat sheet your future self wishes your past self had found sooner. Every command Claude Code understands, laid out in one place, with real examples and plain-English explanations of what actually happens when you run them. No detours, no rabbit holes, no Vim. Bookmark it, print it, tape it to your monitor — your future self (the one who finishes work before dinner) will thank you.
⚠️ Full disclosure: I didn't write this cheat sheet—Claude did. 🤖 I'm just the copy-paste department. 📋 If you trust Claude, we're good. 😎
Table of contents
- Getting started in 60 seconds
- Part 1 — Session commands
- Part 2 — CLI flags
- Part 3 — In-session slash commands
- Part 4 — In-prompt special syntax
- Beginner cheat sheet
Getting started in 60 seconds
# 1. Install
npm install -g @anthropic-ai/claude-code
# 2. Log in
claude auth login
# 3. Open your first session
claude
Inside a session, type /help to see all available commands at any time.
Part 1 — Session commands
These are the commands you run from your terminal (shell) to start, manage, and control Claude Code sessions.
| Command | Syntax | Example | Expected output |
|---|---|---|---|
| Start interactive session | claude |
$ claude |
Claude Code prompt appears, ready for input |
| Start with initial message | claude "query" |
$ claude "explain this project" |
Session opens and Claude immediately responds |
| One-shot mode (no interaction) | claude -p "query" |
$ claude -p "what does this function do" |
Answer printed, Claude exits |
| Pipe file contents to Claude | `cat file \ | claude -p "query"` | `$ cat logs.txt \ |
| Continue last conversation | {% raw %}claude -c
|
$ claude -c |
Previous session reloaded with full history |
| Resume a named session | claude -r "name" "query" |
$ claude -r "auth-refactor" "finish the PR" |
Named session reopens with your message |
| Log in to your account | claude auth login |
$ claude auth login |
Browser opens for auth, then "Logged in" confirmation |
| Log out | claude auth logout |
$ claude auth logout |
"Logged out" confirmation |
| Check login status | claude auth status |
$ claude auth status |
JSON showing login state |
| Update Claude Code | claude update |
$ claude update |
Latest version downloaded and installed |
| Install a specific version | claude install [version] |
$ claude install stable |
Native binary installed |
| View background agents | claude agents |
$ claude agents --json |
Live list of sessions as JSON |
| Stop a background session | claude stop <id> |
$ claude stop 7c5dcf5d |
Session stopped |
| View background session logs | claude logs <id> |
$ claude logs 7c5dcf5d |
Recent output from that session |
| Configure MCP servers | claude mcp |
$ claude mcp |
MCP configuration menu |
| Delete project data | claude project purge [path] |
$ claude project purge ~/work/repo --dry-run |
Preview of what would be deleted |
| Generate a CI/CD token | claude setup-token |
$ claude setup-token |
Long-lived token printed (not saved) |
Part 2 — CLI flags
Flags modify how a command behaves. Append them to any claude command.
| Flag | What it does | Example | Expected output |
|---|---|---|---|
-p / --print
|
Non-interactive mode — Claude answers and exits. Use in scripts. | claude -p "explain main.py" |
Answer printed, process exits with code 0 |
-c / --continue
|
Load the most recent conversation in the current folder. | claude --continue |
Previous session loaded |
-r / --resume
|
Resume a session by name or ID. No value = interactive picker. | claude -r "auth-refactor" |
Named session reopens |
-n / --name
|
Give your session a friendly name for later resumption. | claude -n "my-feature-work" |
Session name shown in title and session list |
-v / --version
|
Print the installed version number. | claude -v |
Version string e.g. 2.1.x
|
--model |
Use a specific model this session (aliases: sonnet, opus). |
claude --model sonnet |
Session runs on the specified model |
--effort |
Set effort level: low, medium, high, xhigh, max. |
claude --effort high |
Session uses the chosen effort level |
--verbose |
Show all intermediate steps and tool calls. | claude --verbose |
Full turn-by-turn output including tool use |
--output-format |
Set output format for -p mode: text, json, stream-json. |
claude -p "query" --output-format json |
Structured JSON with cost, duration, response |
--max-turns |
Limit agentic turns in -p mode. Exits with error if exceeded. |
claude -p --max-turns 3 "fix tests" |
Claude runs at most 3 turns then stops |
--permission-mode |
How Claude handles permissions: default, acceptEdits, plan, auto, bypassPermissions. |
claude --permission-mode plan |
Claude proposes changes before making them |
--add-dir |
Grant Claude read/write access to extra directories. | claude --add-dir ../lib ../apps |
Files in those paths become accessible |
--tools |
Restrict which built-in tools Claude can use. | claude --tools "Read" |
Claude can only read files, not edit or run commands |
--system-prompt |
Replace Claude's default system prompt with your own. | claude --system-prompt "You are a Python expert" |
Claude follows your custom instructions |
--append-system-prompt |
Add extra instructions on top of Claude's default prompt. | claude --append-system-prompt "Always use TypeScript" |
Claude follows defaults + your extra rule |
--bg |
Start session in the background and return to terminal immediately. | claude --bg "investigate flaky test" |
Session ID and control commands printed |
--mcp-config |
Load MCP servers from a JSON config file. | claude --mcp-config ./mcp.json |
MCP servers from that file available in session |
--debug |
Enable debug logging. Optionally filter: "api,mcp". |
claude --debug "api,mcp" |
Detailed debug logs printed |
--bare |
Minimal startup — skips hooks, skills, plugins. Faster for scripts. | claude --bare -p "query" |
Fast response with basic tools only |
--max-budget-usd |
Cap API spend for one -p run. |
claude -p --max-budget-usd 5.00 "refactor auth" |
Claude stops when $5.00 is reached |
--chrome |
Enable Chrome browser integration for web automation. | claude --chrome |
Claude can now control a browser |
--rc / --remote-control
|
Start a session you can also control from the Claude app. | claude --rc "My Project" |
Session appears in Claude app as controllable |
Part 3 — In-session slash commands
Once you are inside a Claude Code session, these / commands control the session itself.
| Command | What it does | Example | Expected output |
|---|---|---|---|
/help |
Show all available slash commands. | /help |
Full list of commands with descriptions |
/clear |
Wipe conversation context entirely. Use between unrelated tasks. | /clear |
Context reset, fresh session starts |
/compact [instructions] |
Compress history to free context space. Optionally focus the summary. | /compact Focus on the API changes |
History summarized with your focus applied |
| `/resume [name\ | id]` | Resume a previous session (interactive picker or by name). | /resume auth-refactor |
/add-dir <path> |
Grant access to an extra directory mid-session. | /add-dir ../shared-lib |
Files in that path become accessible |
/rename <name> |
Rename the current session. | /rename payment-refactor |
Name updated immediately |
/rewind |
Go back to a previous message checkpoint. | /rewind |
Interactive checkpoint picker appears |
/debug |
Run Claude's built-in debug skill on your current code. | /debug |
Claude investigates errors systematically |
/code-review |
Run a structured code review on current changes. | /code-review |
Findings listed by severity |
/simplify |
Simplify the current code for readability. | /simplify |
Simplified version suggested |
/batch |
Process a task across multiple files at once. | /batch |
Claude iterates over files and summarizes results |
Part 4 — In-prompt special syntax
You can use these special characters inside any prompt while in a session.
| Syntax | What it does | Example | Expected output |
|---|---|---|---|
@./path/to/file |
Reference a specific file. Claude reads it and applies your instruction. | Review @./src/auth.ts for bugs |
Claude reads the file and gives a detailed review |
!shell-command |
Run a shell command from inside Claude Code without leaving the session. | !npm test |
Command runs; output shown inline |
Beginner cheat sheet
# Start a session
claude
# Ask a quick question without opening a session
claude -p "explain what a closure is in JavaScript"
# Pipe a file and ask about it
cat server.js | claude -p "find any security issues"
# Come back to your last conversation
claude -c
# Name a session so you can find it later
claude -n "feature-login"
# Resume it later
claude -r "feature-login"
# Inside a session — reference a file
> Review @./src/api.ts and suggest improvements
# Inside a session — run a shell command
> !git status
# Inside a session — compact when context gets long
> /compact
Source: Claude Code official documentation — June 2026