Claude Code CLI cheat sheet

Install and launch#

One native binary. It updates itself in the background.

macOS, Linux, WSL
curl -fsSL https://claude.ai/install.sh | bash
Homebrew
brew install --cask claude-code
Windows PowerShell
irm https://claude.ai/install.ps1 | iex

Launch

Launch
Command	What it does	Copy
`claude`	Start an interactive session
`claude "fix the failing test"`	Start with a first prompt
`claude -c`	Continue the most recent conversation
`claude -r auth-refactor`	Resume a session by name or ID
`claude -p "explain this repo"`	Print one answer and exit (scripts, CI)
`cat error.log \| claude -p "what broke?"`	Pipe input into a one-shot query
`claude --version`	Check the installed version

Everyday commands#

Slash commands go in the Claude Code prompt. Type / to see them all as you go.

Slash commands

Slash commands
Command	What it does	Copy
`/init`	Write a starter CLAUDE.md from your codebase
`/help`	List every command
`/model`	Switch models
`/effort`	Set how hard the model thinks
`/context`	See what’s filling the context window
`/compact`	Summarize the conversation to free up context
`/clear`	Start fresh in the same session
`/resume`	Pick up an earlier session
`/rewind`	Roll back the conversation and file edits to an earlier point
`/btw`	Ask a side question without adding it to the context
`/code-review`	Review pending changes (alias /review); add --fix to apply
`/usage`	Session cost, plan limits, activity (alias /cost)
`/memory`	Edit the CLAUDE.md files in play
`/powerup`	Interactive lessons on features you haven’t tried
`/doctor`	Diagnose your install and settings

Keyboard shortcuts

Keyboard shortcuts
Keys	What it does
`Shift+Tab`	Cycle permission modes
`Esc`	Interrupt Claude or close a dialog
`EscEsc`	Clear the draft, or open rewind when it’s empty
`Ctrl+R`	Search prompt history
`Ctrl+B`	Send a running task to the background
`Ctrl+O`	Open the full transcript
`Ctrl+G`	Edit the prompt in your $EDITOR
`Option+P`	Switch model (Alt+P on Windows/Linux)
`Option+T`	Toggle extended thinking (Alt+T)
`Option+O`	Toggle fast mode (Alt+O)

Permission modes#

How much Claude may do without asking. Press Shift+Tab in a session to cycle, or start in a mode with --permission-mode.

Permission modes
Mode	What it does	Use when
Manual`--permission-mode default`	Asks before edits and commands	Learning a codebase, or anything risky
Edit automatically`--permission-mode acceptEdits`	Edits files and runs basic file commands (mkdir, mv, cp) without asking; asks for everything else	You trust the edits but want to see commands
Plan`--permission-mode plan`	Reads and plans; no edits until you approve the plan	Before a big change, or to understand code
Auto`--permission-mode auto`	A classifier reviews each action and blocks risky ones; no prompts for the rest	Well-scoped work with a good CLAUDE.md
Don’t ask`--permission-mode dontAsk`Flag or settings only	Runs only pre-approved tools; denies anything that would prompt	Scripts and CI with an allowlist
Bypass permissionsDanger`--dangerously-skip-permissions`In the cycle only once enabled	No prompts, no checks. Deny rules still apply.	Throwaway containers and VMs only

Auto mode

“Seatbelts, not a force field.”

How it works: Before each action runs, a classifier (Sonnet 5 by default) checks it against your request and blocks anything destructive or out of scope. It’s a safety net, not a sandbox.
Who gets it: All plans, on supported models. On by default for Team and Enterprise; admins can turn it off. Recent versions start interactive sessions in auto mode.

Interactive
claude --permission-mode auto
Headless
claude -p "run the tests and fix failures" --permission-mode auto

CLAUDE.md#

A markdown file Claude reads at the start of every session: what the project is, how to build it, and what not to touch. Run /init to draft one.

Where it lives

./CLAUDE.md: Project, committed with the repo (or ./.claude/CLAUDE.md)
./CLAUDE.local.md: Project, just for you (gitignored)
~/.claude/CLAUDE.md: You, across every project
.claude/rules/*.md: Topic rules, optionally scoped to paths

Pull other files in with @path/to/file. A CLAUDE.md in a subfolder loads when Claude works there.

What makes one good

Be specific: “Use snake_case for database columns” beats “follow our conventions.”
Say what not to do: Constraints prevent more mistakes than instructions do.
Keep it current: When Claude gets something wrong twice, the fix belongs here.

CLAUDE.md

# CLAUDE.md

## Project Overview
What this project does, in two sentences.

## Build & Run
```bash
npm run dev        # Development
npm run build      # Production build
npm test           # Run tests
```

## Architecture
Key patterns, file layout, data flow.

## Conventions
- Naming, imports, file organization
- See @docs/style-guide.md

## What NOT To Do
- Anti-patterns specific to this project
- Files and directories never to touch

Skills#

A folder with a SKILL.md: instructions, and optionally scripts and reference files. Claude loads one when its description matches the task, or you run it as /skill-name.

.claude/skills/release-notes/SKILL.md

---
name: release-notes
description: Draft release notes from merged PRs. Use when the user asks for a changelog or release notes.
disable-model-invocation: false
---

# Release notes

1. List PRs merged since the last tag.
2. Group them: features, fixes, chores.
3. Write one plain sentence per change.

Where they live

.claude/skills/<name>/SKILL.md: This project
~/.claude/skills/<name>/SKILL.md: Every project

The description is the trigger: Claude decides whether to load a skill from its description alone, so say when to use it.
Skills are slash commands: Every skill can be run as /name. Add disable-model-invocation: true for skills you only want to run by hand.
An open standard: Skills follow the Agent Skills format, so the same folder works in other tools that support it.

Find more

Plugins#

A plugin bundles skills, commands, hooks, agents, and MCP servers into one install. Browse with /plugin. These two are the author’s picks.

Superpowers

By Jesse Vincent and Prime Radiant · github.com/obra/superpowers

Makes Claude brainstorm, plan, and test before it writes code, then review before it ships.

/plugin install superpowers@claude-plugins-official

The workflow it enforces

BrainstormPin down what you actually want
Design docWrite the spec down
WorktreeIsolate the work on a branch
PlanBreak the spec into small steps
TDDFailing test first, then the code
SubagentsHand steps to fresh agents
ReviewCheck the work against the plan
FinishMerge, PR, or clean up

Skills you’ll see it use

brainstorming: Questions before code
writing-plans: Specs broken into bite-size tasks
test-driven-development: Red, green, refactor
subagent-driven-development: One fresh agent per task
requesting-code-review: Review against the plan
finishing-a-development-branch: Merge or PR, then clean up

Impeccable

By Paul Bakaus · impeccable.style

Gives Claude a design vocabulary: critique a page, audit it for accessibility, then polish, simplify, or push it bolder.

/plugin marketplace add pbakaus/impeccable

/plugin install impeccable@impeccable

Commands worth starting with

Commands worth starting with
Command	What it does	Copy
`/impeccable critique`	Scored design review with priority fixes
`/impeccable audit`	Accessibility, performance, responsive checks
`/impeccable polish`	Final pass before shipping

This page’s redesign started with /impeccable critique. It scored the old version 19 out of 40.

MCP servers#

The Model Context Protocol connects Claude to tools and data outside the repo: GitHub, browsers, databases, docs. Remote servers use HTTP; local ones run as a process (stdio).

Add a server

Remote (HTTP)
claude mcp add --transport http my-server https://api.example.com/mcp
Local process (stdio)
claude mcp add --transport stdio my-server -- npx -y my-mcp-server
From JSON
claude mcp add-json my-server '{"type":"stdio","command":"npx","args":["-y","my-mcp-server"]}'

Scopes

Scopes
Scope	Stored in	Applies to
Local`(default)`	`~/.claude.json`	You, in this project only
Project`--scope project`	`.mcp.json`	Everyone on the repo (commit it)
User`--scope user`	`~/.claude.json`	You, in every project

Popular servers

GitHub · Issues, PRs, code search (GitHub’s hosted server)
claude mcp add --transport http github https://api.githubcopilot.com/mcp/ --header "Authorization: Bearer YOUR_GITHUB_PAT"
Playwright · Drive a real browser
claude mcp add playwright -- npx @playwright/mcp@latest
Context7 · Current docs for the libraries you use
claude mcp add --transport stdio context7 -- npx -y @upstash/context7-mcp
Brave Search · Web search
claude mcp add --transport stdio --env BRAVE_API_KEY=YOUR_KEY brave -- npx -y @brave/brave-search-mcp-server
Filesystem · Files outside the project
claude mcp add --transport stdio filesystem -- npx -y @modelcontextprotocol/server-filesystem ~/notes
Sequential Thinking · Step-by-step reasoning scratchpad
claude mcp add --transport stdio thinking -- npx -y @modelcontextprotocol/server-sequential-thinking

Manage

Manage
Command	What it does	Copy
`claude mcp list`	List configured servers
`claude mcp get <name>`	Show one server’s config
`claude mcp remove <name>`	Remove a server
`/mcp`	Status and sign-in, inside a session

Build your own (Python, FastMCP)

server.py

from fastmcp import FastMCP

mcp = FastMCP("my-tools")

@mcp.tool
def search_docs(query: str) -> str:
    """Search project documentation."""
    return f"No results for {query!r} yet."

if __name__ == "__main__":
    mcp.run()

Spec and SDKs: modelcontextprotocol.io

Hooks#

Commands that run on Claude Code events, every time, whatever the model decides. Configure them in .claude/settings.json or with /hooks.

Events you’ll use most

Events you’ll use most
Event	Fires	Copy
`SessionStart`	When a session starts or resumes
`UserPromptSubmit`	Before Claude sees your prompt
`PreToolUse`	Before a tool runs (can block it)
`PermissionRequest`	When a permission prompt would appear
`PostToolUse`	After a tool succeeds
`Notification`	When Claude needs your attention
`Stop`	When Claude finishes responding
`PreCompact`	Before the context is compacted

There are about 30 more, from SubagentStop to FileChanged.Full list

Handlers can be a shell command, an HTTP call, an MCP tool, or a prompt for a model to judge.

Format every edit

Runs Prettier on each file Claude writes. The file path arrives as JSON on stdin.

.claude/settings.json

{
  "hooks": {
    "PostToolUse": [
      {
        "matcher": "Write|Edit",
        "hooks": [
          {
            "type": "command",
            "command": "jq -r '.tool_input.file_path' | xargs npx prettier --write"
          }
        ]
      }
    ]
  }
}

Tell me when it’s done

A macOS notification each time Claude stops.

~/.claude/settings.json

{
  "hooks": {
    "Stop": [
      {
        "hooks": [
          {
            "type": "command",
            "command": "osascript -e 'display notification \"Claude is done\" with title \"Claude Code\"'"
          }
        ]
      }
    ]
  }
}

Subagents#

Claude can hand work to subagents. Each gets its own context window, tools, and instructions, and returns only a summary, so your main context stays clean.

Built in

Explore: Fast, read-only search across the codebase
Plan: Research for plan mode
general-purpose: Multi-step tasks that need every tool

Write your own

A markdown file in .claude/agents/ (this project) or ~/.claude/agents/ (every project). Ask Claude to write one.

.claude/agents/test-runner.md

---
name: test-runner
description: Runs the test suite and fixes failures. Use after code changes.
tools: Read, Edit, Bash
model: sonnet
---

Run the tests. For each failure, find the cause,
fix it, and re-run until the suite passes.
Report what you changed.

Links#

The official docs, the repos mentioned above, and a skill from the talk.

Download

prd-writer
Writes PRDs precise enough for Claude Code to build from unattended: structured sections, quality gates, stack defaults.
Download skill

Claude Code CLI, on one page

Install and launch#

Launch

Everyday commands#

Slash commands

Keyboard shortcuts

Permission modes#

Auto mode

CLAUDE.md#

Where it lives

What makes one good

Skills#

Where they live

Find more

Plugins#

Superpowers

The workflow it enforces

Skills you’ll see it use

Impeccable

Commands worth starting with

MCP servers#

Add a server

Scopes

Popular servers

Manage

Build your own (Python, FastMCP)

Hooks#

Events you’ll use most

Format every edit

Tell me when it’s done

Subagents#

Built in

Write your own

Links#

Download

Read more

✻Claude Code CLI, on one page

Launch

Slash commands

Keyboard shortcuts

Auto mode

Where it lives

What makes one good

Where they live

Find more

The workflow it enforces

Skills you’ll see it use

Commands worth starting with

Add a server

Scopes

Popular servers

Manage

Build your own (Python, FastMCP)

Events you’ll use most

Format every edit

Tell me when it’s done

Built in

Write your own

Download

Read more

Claude Code CLI, on one page