Skip to content

Latest commit

 

History

History
2078 lines (1717 loc) · 145 KB

claude-code-grok.md

File metadata and controls

2078 lines (1717 loc) · 145 KB
title Claude Code
subtitle Your first AI coding partner, from installation to everyday mastery
volume 1
edition_label Grok Edition
edition_note Concise edition, revised and fact-checked by Grok Bot in GrokAllManualsV1.
edition October 2026
accent #C8603C
accent2 #F2B84B
motif orbits
blurb Claude Code is an AI assistant from Anthropic that can read your project, write and edit files, run commands, and explain everything it does in plain English. You talk to it the way you'd talk to a helpful colleague, and it does the work. This manual assumes you have **never used a terminal before**. Each chapter builds on the one before it. By the end you'll have installed Claude Code, built and changed real files with it, learned to control what it is allowed to do, and set it up to remember how you like to work.

Welcome to Claude Code

Claude Code is an AI coding agent. That means it can do more than answer questions: it can open the files in your project, change them, run programs, check its own work, and keep going until a task is done. You stay in charge the whole time.

:::summary In this chapter

  • What Claude Code is and what it's good for
  • The different places you can use it
  • What you need before you start :::

What is an "agent"?

You have probably used a chatbot such as the Claude app or ChatGPT. You type a question and it types back an answer. If it suggests code, you copy it, paste it into a file, and run it.

An agent closes that loop. When you ask Claude Code to "add a contact form to my website," it will:

  1. Look through your project to understand how it's built.
  2. Decide which files need to change.
  3. Edit those files (asking your permission first, unless you've told it not to).
  4. Run commands, for example starting your website, to check that the change works.
  5. Tell you what it did, in plain language.

This cycle of look → plan → act → check is called the agentic loop. It's what makes Claude Code feel less like a search engine and more like a teammate.

What can you do with it?

You don't need to be a programmer to get value from Claude Code. People use it to:

  • Build things from a description. "Make me a one-page website for my bakery with our hours and a menu."
  • Understand unfamiliar code. "Explain what this project does, as if I'm new to coding."
  • Fix problems. "When I run the app I get this error. Find out why and fix it."
  • Automate chores. "Rename all these photos by the date they were taken."
  • Work with Git and GitHub. "Commit my changes with a clear message and open a pull request."
  • Write documents and data files, not just code: reports, spreadsheets, scripts, and more.

Tip

You can ask Claude Code about itself. Questions like "what can Claude Code do?" or "how do I add an MCP server?" work, because it has its own documentation built in.

Where Claude Code runs

Claude Code is one product with several "front doors." They share the same brain and the same settings, so what you learn in this manual works everywhere.

Surface What it looks like Best for
Terminal (CLI) A text window where you type The full feature set; the focus of this manual
Desktop app The Claude app's Code tab People who prefer buttons and windows over typing commands
VS Code / Cursor extension A chat panel inside your code editor Seeing edits side-by-side with your files
JetBrains plugin Runs inside IntelliJ, PyCharm, etc. Developers already using those editors
Web claude.ai/code Running tasks in the cloud, even from a phone
Slack, GitHub Actions Mention Claude in a chat or pull request Teams and automation

CLI stands for command-line interface: a program you control by typing text commands instead of clicking.

What you need

:::checklist

  • A computer running macOS 13+, Windows 10 (1809+), or a recent Linux (Ubuntu 20.04+, Debian 10+), with at least 4 GB of RAM
  • An internet connection
  • A paid Claude account: Pro, Max, Team, or Enterprise, or an Anthropic Console account that pays per use
  • A folder to work in (it can be empty, so you don't need an existing project) :::

Note

The free Claude plan does not include Claude Code. The cheapest way in is the Pro plan. See the pricing table in Chapter 3.

Terminal Basics in Ten Minutes

The terminal looks intimidating, but you only need a handful of ideas to use Claude Code comfortably. Think of it as a very literal text message conversation with your computer.

:::summary In this chapter

  • How to open a terminal on Mac, Windows, and Linux
  • Folders, paths, and the "current directory"
  • The five commands you'll actually use :::

Opening a terminal

:::steps Open your terminal

  1. On a Mac. Press Cmd + Space, type Terminal, and press Enter.
  2. On Windows. Press the Windows key, type PowerShell, and open Windows PowerShell (or Terminal, which opens PowerShell by default).
  3. On Linux. Press Ctrl + Alt + T on most systems, or search your apps for "Terminal." :::

You'll see a line ending in something like %, $, or >. That's the prompt: the terminal waiting for you to type. When this manual shows a command, type it after the prompt and press Enter.

Tip

On Windows, the prompt starts with PS C:\ when you're in PowerShell and just C:\ when you're in the older Command Prompt (CMD). Claude Code works in both, but the install commands differ.

Folders and paths

Everything on your computer lives in folders (also called directories). A path is the address of a folder or file, like /Users/sam/projects/bakery on a Mac or C:\Users\Sam\projects\bakery on Windows.

The terminal is always "standing in" one folder, called the current directory or working directory. Claude Code works on the files in whatever folder you start it from. That's how it knows which project you mean.

The ~ symbol is shorthand for your home folder (for example /Users/sam).

The five commands you need

Command What it does Example
pwd Show which folder you're in ("print working directory") pwd
ls List the files here (dir in CMD) ls
cd Change into a folder cd projects/bakery
cd .. Go up one folder cd ..
mkdir Make a new folder mkdir bakery

Here's how to make a practice folder and step inside it:

cd ~
mkdir my-first-project
cd my-first-project
pwd

Tip

Press Tab while typing a folder name and the terminal finishes it for you. Press the ↑ arrow to bring back the last command you typed.

Copying and pasting in a terminal

  • Mac: Cmd + C and Cmd + V, as usual.
  • Windows Terminal / PowerShell: Ctrl + V pastes; right-click also pastes.
  • Linux: usually Ctrl + Shift + V.

In a terminal, Ctrl + C usually means stop the running program, not copy.

Install and Sign In

Installation takes about two minutes. Anthropic's recommended method is the native installer, which also keeps Claude Code up to date automatically.

:::summary In this chapter

  • Installing on macOS, Linux, and Windows
  • Choosing a plan and logging in
  • Checking that everything works :::

Install with the native installer

:::steps Install Claude Code

  1. Open a terminal. See Chapter 2 if you're not sure how.

  2. Paste the command for your system and press Enter.

    macOS, Linux, or WSL:

    curl -fsSL https://claude.ai/install.sh | bash

    Windows PowerShell:

    irm https://claude.ai/install.ps1 | iex

    Windows Command Prompt (CMD):

    curl -fsSL https://claude.ai/install.cmd -o install.cmd && install.cmd && del install.cmd
  3. Wait. The installer shows no progress bar while it downloads. That's normal.

  4. Open a new terminal window so it picks up the new program.

  5. Check the install. Type claude --version. You should see a version number followed by (Claude Code). :::

[!WARNING] "command not found"? If your terminal says claude isn't found or isn't recognized, the install folder isn't on your PATH yet. (The PATH is the list of folders your terminal searches for programs.) Close every terminal window and open a fresh one first. If it still fails, run the installer again and read its last few lines. It prints the exact fix.

Other ways to install

Method Command Updates
Homebrew (Mac) brew install --cask claude-code Manual: brew upgrade claude-code
WinGet (Windows) winget install Anthropic.ClaudeCode Manual: winget upgrade Anthropic.ClaudeCode
npm (needs Node.js 22+) npm install -g @anthropic-ai/claude-code Manual
apt / dnf / apk (Linux) See the setup page in the official docs Through your system updates

Caution

Never install with sudo npm install -g. Running npm as an administrator causes permission problems and is a security risk.

Windows notes

  • You do not need to run the installer as Administrator.
  • Installing Git for Windows (from git-scm.com) is recommended. It gives Claude Code a Bash shell to run commands with. Without it, Claude Code uses PowerShell instead, which also works.
  • If you already use WSL (Windows Subsystem for Linux), install from inside your WSL terminal using the macOS/Linux command.

Choose a plan

Claude Code needs a paid account. Prices below are from claude.com/pricing in October 2026:

Plan Price Who it's for
Free $0 Chat only, no Claude Code
Pro $20/month ($17/month billed yearly) Learning and light daily use
Max 5x $100/month Heavy daily use: 5× Pro's usage
Max 20x $200/month All-day use: 20× Pro's usage
Team $25/seat/month Standard, $125 Premium (less yearly) Small teams, 2–150 people
Enterprise Custom Large organizations
Console (API) Pay per use Developers who prefer metered billing

Tip

Start with Pro. If you regularly hit your usage limit, upgrade to Max later. Inside Claude Code, /usage shows how much of your limit you've used.

Log in for the first time

:::steps Sign in

  1. Go to your project folder. For example: cd ~/my-first-project.
  2. Start Claude Code. Type claude and press Enter.
  3. Pick a theme. Use the arrow keys to choose light or dark.
  4. Choose how to log in. Pick your Claude subscription (Pro/Max/Team/Enterprise) or Anthropic Console.
  5. Approve in your browser. A browser window opens; click Authorize. Return to the terminal.
  6. Trust the folder. Claude Code asks whether you trust the files in this folder. For your own projects, choose Yes. :::

You only log in once; Claude Code remembers your credentials. To switch accounts later, type /login inside Claude Code.

Health check and updates

  • claude doctor checks your installation and settings and suggests fixes. Run it whenever something feels off.
  • Native installs update themselves in the background. To update right now: claude update.
  • Prefer slower, battle-tested releases? Type /config and set Auto-update channel to stable (about a week behind, skipping releases with major bugs).

Your First Session

Let's build something. In this chapter you'll start Claude Code, ask it questions, and have it create and change real files.

:::summary In this chapter

  • What the Claude Code screen shows you
  • Asking questions and requesting changes
  • Approving actions, interrupting, and undoing :::

The Claude Code screen

When you run claude, you'll see:

  • A welcome panel showing the version, the AI model in use, and your current folder.
  • A prompt box where you type, marked with >.
  • A status bar underneath showing the current permission mode (more on this in Chapter 6).

Type your request in plain English and press Enter. There's no special syntax to learn.

Tip

To write several lines before sending, press Shift + Enter for a new line (or type \ then Enter, which works in every terminal).

Build a small website

:::steps Your first project

  1. Start in an empty folder.
    mkdir ~/bakery-site
    cd ~/bakery-site
    claude
  2. Describe what you want.
    Create a simple one-page website for a bakery called "Rise & Shine".
    Include our hours (Tue–Sun, 7am–3pm), a short menu with prices,
    and a contact section. Keep it to plain HTML and CSS.
    
  3. Watch it plan and work. Claude explains what it's about to do, then creates files such as index.html and styles.css.
  4. Approve when asked. If Claude asks permission to create a file, choose Yes. (Chapter 6 explains how to approve whole categories at once.)
  5. Open the result. Ask: open the website in my browser. Or open index.html from your file manager.
  6. Change it. Try: make the colors warmer and add a photo placeholder at the top. :::

That's the whole rhythm: describe, review, refine.

Ask questions about existing code

Claude Code shines when you point it at a project someone else wrote. Start it inside the project folder and try:

what does this project do?
explain the folder structure like I'm new to coding
where is the main entry point, and what happens when the app starts?

You don't need to tell Claude which files to read. It looks around by itself. If you want to point at a file, type @ and start typing its name; Claude Code autocompletes the path.

Approving, interrupting, and undoing

Permission prompts. Before Claude edits a file or runs a command it isn't sure about, it may show what it intends to do and offer choices like:

  • Yes: allow this one action.
  • Yes, and don't ask again: allow this kind of action from now on in this project.
  • No: refuse, and optionally tell Claude what to do instead.

Interrupting. Press Esc at any time to stop Claude mid-task. It keeps what it has done so far, and you can redirect it: "Stop. Use blue instead of green."

Undoing. Claude Code saves a checkpoint before each change. Press Esc twice on an empty prompt (or type /rewind) to roll the conversation and/or your files back to an earlier point.

Warning

Checkpoints cover file edits Claude makes. They can't undo things outside your computer, like a sent email, a deleted cloud file, or a database change. For real safety, use Git (Chapter 6 and the GitHub manual in this series).

Ending and resuming

To… Do this
Leave Claude Code Type /exit, or press Ctrl + D twice
Start fresh in the same window /clear
Continue your last conversation claude -c (from the terminal)
Pick an older conversation claude -r, or /resume inside Claude Code
Name a session to find it later /rename bakery-site

Talking to Claude Effectively

Claude Code is capable, but it isn't a mind reader. A little care in how you ask makes a huge difference in what you get back.

:::summary In this chapter

  • The habits that produce great results
  • Plan mode: think first, then build
  • Sharing screenshots, errors, and files :::

Five habits of great prompts

  1. Be specific. Instead of "fix the bug," say "users see a blank screen after entering a wrong password, so fix that."
  2. Give the goal, not just the task. "I want customers to order ahead online" helps Claude make better choices than "add a form."
  3. Break big jobs into steps. Number them, and Claude will work through them in order.
  4. Let Claude explore first. "Read the code and explain how login works before changing anything."
  5. Say how you'll judge success. "When you're done, run the tests and show me they pass."
1. Add a "Order ahead" page with a form for name, item, and pickup time.
2. Save orders to a file called orders.csv.
3. Add a link to the new page in the site's menu.
4. Open the site and confirm the form works before you finish.

Plan mode: look before you leap

For anything bigger than a small tweak, ask Claude to plan first. In plan mode Claude researches and writes a step-by-step plan but won't edit your files until you approve it.

  • Press Shift + Tab until the status bar says plan mode on, or
  • Start your message with /plan, for example: /plan add online ordering.

When the plan appears you can approve it, ask for changes ("No, keep planning, and don't use a database"), or press Ctrl + G to edit the plan yourself in a text editor.

Tip

Plan mode is the single best habit for beginners. Reading the plan teaches you how the project works, and it catches misunderstandings before any files change.

Share what you see

  • Error messages: copy and paste them straight into the prompt. Include the whole thing.
  • Screenshots: paste an image with Ctrl + V (Cmd + V in iTerm2, Alt + V on Windows). Claude can see it.
  • Files: type @ followed by part of a file name.
  • A command's output: start a line with ! to run a shell command yourself, for example !npm test. Claude sees the result and responds to it.

Ask a side question

Want to ask something without derailing the current task? Use /btw:

/btw what's the difference between HTML and CSS?

The answer appears without being added to the main conversation.

Keep conversations focused

Claude can hold a lot in mind at once, but that space, the context window, is not unlimited. Long, wandering sessions get slower and cost more of your usage.

  • Use /clear when you switch to an unrelated task.
  • Use /compact to summarize a long conversation and free up space while staying on the same task.
  • Use /context to see how full the context window is.

Permissions and Safety

Claude Code can edit files and run commands on your computer, so it comes with a permission system: rules for what it may do without asking. Understanding it lets you work faster and safer.

:::summary In this chapter

  • The six permission modes and how to switch
  • Allow, ask, and deny rules
  • Safety nets: checkpoints, Git, and the sandbox :::

Permission modes

A permission mode sets how much Claude can do before checking with you. Press Shift + Tab to cycle modes during a session. The status bar shows the active one.

Mode Runs without asking Good for
Manual (default) Reading files only Sensitive work; learning what Claude does
Accept edits (acceptEdits) Reading, editing files, and simple file commands like mkdir, mv, cp Iterating on code you'll review afterwards
Plan (plan) Reading and exploring; no edits until you approve a plan Thinking before building
Auto (auto) Almost everything, with a safety classifier checking each action in the background Long tasks with few interruptions
Don't ask (dontAsk) Only pre-approved tools; everything else is refused Scripts and automation
Bypass (bypassPermissions) Everything, no checks Disposable containers and virtual machines only

Auto mode is the starting mode for new terminal and VS Code sessions on recent versions of Claude Code. A second AI model (the classifier) reviews each risky action, such as running a script from the internet, force-pushing code, deploying, or deleting files that existed before the session, and blocks anything that goes beyond what you asked for.

Important

You can tell auto mode about limits in plain English: "don't push anything until I've reviewed it." The classifier treats that as a rule for the rest of the conversation. For a hard guarantee, add a deny rule (below).

To start a session in a particular mode:

claude --permission-mode plan

Allow, ask, and deny rules

Rules let you pre-approve safe actions and forbid dangerous ones. Type /permissions to see and edit them, or write them in a settings file.

  • Allow: run without asking.
  • Ask: always ask first, even in auto mode.
  • Deny: never allowed, in any mode.

Deny rules win over ask rules, which win over allow rules.

{
  "permissions": {
    "allow": [
      "Bash(npm run *)",
      "Bash(git commit *)"
    ],
    "ask": [
      "Bash(git push *)"
    ],
    "deny": [
      "Read(./.env)",
      "Bash(rm -rf *)"
    ]
  }
}

The * matches anything, so Bash(npm run *) covers npm run build, npm run test, and so on. Put the * after the sub-command: Bash(git commit *) is narrow, but Bash(git *) allows every Git command.

Tip

Tired of approving the same harmless commands? Type /fewer-permission-prompts. Claude scans your recent sessions and proposes a safe allow-list for read-only commands you use often.

Where settings live

Claude Code reads settings from several files. When the same setting appears in more than one place, the higher one in this list wins:

Priority File Applies to
1 (highest) Managed settings from your organization Everyone in your company
2 claude --settings … on the command line This session
3 .claude/settings.local.json You, in this project (not shared)
4 .claude/settings.json Everyone working in this project
5 ~/.claude/settings.json You, in every project

~ means your home folder. A file inside .claude/ lives in the project folder.

Safety nets

  • Checkpoints: /rewind (or Esc Esc) restores earlier file versions.
  • Git: commit your work before big changes so you can always get back. Ask Claude: "commit everything with a clear message before you start."
  • Protected files: Claude never silently edits things like .git, .bashrc, or its own .claude settings. It always asks, except in bypass mode.
  • Sandbox: type /sandbox (macOS, Linux, WSL2) to run Claude's shell commands in an isolated box with limited file and network access.

[!CAUTION] Bypass mode is for throwaway machines --dangerously-skip-permissions turns off every check. A bad instruction hidden in a web page or file (a prompt injection) could then make Claude do real damage. Only use it inside a container or virtual machine you're happy to wipe. The Maximum Autonomy guide in this series covers how to set that up.

Memory: Teaching Claude Your Project

Every session starts with a blank slate. Two features carry knowledge from one session to the next: CLAUDE.md files, which you write, and auto memory, which Claude writes.

:::summary In this chapter

  • What a CLAUDE.md file is and where to put it
  • Generating one with /init
  • What belongs in it, with an example :::

CLAUDE.md files

A CLAUDE.md file is a plain-text note, written in Markdown (a simple way to format text with # headings and - bullets), that Claude reads at the start of every session. Use it for anything you'd otherwise re-explain each time: how to run the project, house style, "always do X" rules.

Where File Who it's for
Your whole computer ~/.claude/CLAUDE.md Your personal preferences, every project
A project (shared) ./CLAUDE.md or ./.claude/CLAUDE.md Everyone on the project, checked into Git
A project (just you) ./CLAUDE.local.md Your private notes; add it to .gitignore

Create one with /init

Inside a project, type:

/init

Claude studies the project and writes a starter CLAUDE.md with build commands, test instructions, and conventions it finds. If one already exists, it suggests improvements instead of overwriting.

What to put in it

# Rise & Shine bakery website

## How to preview
- Open index.html in a browser. No build step.

## Style
- Plain HTML and CSS only. No frameworks.
- Colors: warm cream (#FFF6E9) and cinnamon (#B5562B).
- Write friendly, short copy. No jargon.

## Rules
- Keep every page working on phones.
- Ask before deleting any image.

Tips from Anthropic's guidance:

  • Keep it short: aim for under about 200 lines. Long files get followed less reliably.
  • Be concrete: "Use 2-space indentation" beats "format code nicely."
  • Add a line when Claude makes the same mistake twice.
  • Import other files with @, e.g. See @README.md for an overview.

Note

CLAUDE.md is guidance, not a lock. To truly block an action, use a deny rule (Chapter 6) or a hook (Chapter 9).

Auto memory

As you work, Claude can note your corrections and preferences ("Sam prefers short commit messages") in its own memory files, which load automatically next time. Type /memory to view or edit them, open your CLAUDE.md files, or turn auto memory on or off.

Everyday Commands and Shortcuts

You don't need to memorize these. Type / at any time to see a searchable menu of everything available. Here are the ones you'll use most.

Slash commands

Command What it does
/help Show help and available commands
/init Create a CLAUDE.md for this project
/clear Start a new, empty conversation
/compact Summarize the conversation to free up space
/context Show how full the context window is
/model Switch AI model (e.g. opus, sonnet, haiku)
/plan Enter plan mode, optionally with a task
/permissions View and edit allow / ask / deny rules
/rewind Roll back the conversation and/or files
/resume Reopen an earlier conversation
/diff Review all changes made so far
/review Ask Claude to review your changes for bugs
/memory Edit CLAUDE.md files and auto memory
/mcp Manage connected MCP servers (Chapter 9)
/usage See plan limits and usage (/cost works too)
/config Open settings: theme, model, editor mode…
/doctor Diagnose and fix setup problems
/add-dir Let Claude work in an extra folder
/export Save the conversation as text
/btw Ask a side question
/exit Quit

Keyboard shortcuts

Keys Action
Esc Stop Claude mid-task
Esc Esc Clear your draft; on an empty prompt, open rewind
Shift + Tab Cycle permission modes
Shift + Enter New line without sending
↑ / ↓ Scroll through previous prompts
Ctrl + R Search prompt history
Ctrl + C Interrupt; on an empty prompt, press twice to exit
Ctrl + O Show the detailed transcript of what Claude did
Ctrl + G Edit your prompt (or a plan) in a text editor
Ctrl + V Paste an image
Ctrl + B Send a long-running command to the background
? Show the shortcut help panel (on an empty prompt)

Starting Claude Code from the terminal

Command What it does
claude Start an interactive session
claude "fix the build error" Start with a first request
claude -c Continue the most recent conversation here
claude -r Choose an older conversation to resume
claude -p "summarize README.md" Answer once and exit (handy in scripts)
claude --model opus Use a specific model this session
claude --permission-mode plan Start in a specific permission mode
claude -w new-feature Work in an isolated copy of the repo (a Git worktree)
claude update Update now
claude doctor Check installation health

Extending Claude Code

Out of the box, Claude Code can work with files, run commands, and search the web. You can extend it in five ways. Each has its own job:

Feature In one sentence Lives in
MCP servers Connect Claude to outside tools: GitHub, Notion, databases, browsers claude mcp add …
Skills Save a reusable set of instructions as a /command .claude/skills/<name>/SKILL.md
Subagents Specialist helpers with their own instructions and tools .claude/agents/<name>.md
Hooks Run your own command automatically when something happens "hooks" in settings.json
Plugins Bundles of the above you can install in one step /plugin

MCP: connect to other apps

MCP (Model Context Protocol) is an open standard that lets AI tools plug into other software. An MCP server is a small connector that gives Claude new abilities, such as reading your Notion pages or creating GitHub issues. (This series has a whole manual on MCP.)

Most modern services offer a remote server: a web address. Add one like this:

claude mcp add --transport http notion https://mcp.notion.com/mcp

Then type /mcp inside Claude Code, select the server, and sign in through your browser.

Some servers run as a program on your own computer (a local or stdio server). Everything after the -- is the command that starts it:

claude mcp add --env API_KEY=your-key myserver -- npx -y @example/mcp-server
Command What it does
claude mcp list List servers and whether they're connected
claude mcp get <name> Show details for one server
claude mcp remove <name> Remove a server
--scope user Make a server available in all your projects
--scope project Save it in .mcp.json to share with your team

Warning

Only install MCP servers you trust. A server can see what Claude sends it and can return text that tries to trick Claude. Prefer official servers published by the company whose product you're connecting.

Skills: your own slash commands

A skill is a folder containing a SKILL.md file of instructions. Claude uses it automatically when it fits the situation, or you run it by name.

---
description: Writes a friendly weekly progress update from recent Git commits. Use when asked for a status update or weekly summary.
---

Look at the Git commits from the last 7 days. Write a short update with:
- three bullet points of what got done
- one bullet of what's next
Use plain language a non-programmer would understand.

Now /weekly-update works in every project. Put the folder in a project's .claude/skills/ instead to share it with your team.

Subagents: specialist helpers

A subagent is a helper Claude can hand work to. Each has its own instructions, its own allowed tools, and its own fresh context, so a big research job doesn't clutter your main conversation. The easiest way to make one is to ask:

Create a project subagent called "copy-editor" that checks my website text
for spelling, tone, and reading level. It should only be able to read files.

Claude saves it as .claude/agents/copy-editor.md. Claude Code also ships with built-in subagents, such as Explore (fast, read-only searching) and Plan.

Hooks: automatic actions

A hook runs a command of yours at a specific moment, for example after every file edit or when Claude needs your attention. Unlike CLAUDE.md, hooks always run. This example shows a desktop notification on a Mac whenever Claude is waiting for you:

{
  "hooks": {
    "Notification": [
      {
        "matcher": "",
        "hooks": [
          {
            "type": "command",
            "command": "osascript -e 'display notification \"Claude needs you\" with title \"Claude Code\"'"
          }
        ]
      }
    ]
  }
}

Type /hooks to see what's configured. You can also just describe the hook you want and ask Claude to set it up.

Plugins: install bundles

A plugin packages skills, subagents, hooks, and MCP servers together. Type /plugin to browse and install from marketplaces (catalogs of plugins). Anthropic's official marketplace is built in.

Beyond the Terminal

Everything you've learned carries over to Claude Code's other surfaces.

The desktop app

Download the Claude app from claude.com/download, sign in, and open the Code tab. You get the same Claude Code with a graphical interface: a folder picker, a visual diff (a side-by-side view of what changed), a permission-mode selector next to the send button, and the ability to run several sessions in parallel.

VS Code and Cursor

In VS Code or Cursor, open the Extensions view (Ctrl/Cmd + Shift + X), search for Claude Code, and click Install. A Claude panel appears in the sidebar. You can review proposed changes as inline diffs and @-mention the lines you've selected.

Claude Code on the web and phone

At claude.ai/code, or in the Code tab of the Claude mobile app, you can start a cloud session: Claude works on a copy of your GitHub repository on Anthropic's computers, so it keeps going even when your laptop is closed. When it's done, it pushes a branch you can review. Cloud sessions are available on Pro, Max, Team, and eligible Enterprise seats, and need your GitHub account connected.

From the terminal, claude --cloud "fix the login bug" hands a task to the cloud, and claude --teleport pulls a cloud session back to your computer.

Scripts and automation

claude -p ("print mode") answers once and exits, which makes Claude Code usable inside scripts:

cat error.log | claude -p "explain this error in one paragraph"

For GitHub, type /install-github-app to let your team mention @claude in issues and pull requests to get work done automatically.

Costs and Limits

How usage works

On a subscription (Pro, Max, Team), Claude Code use is included up to your plan's limits, which reset over time. On a Console account you pay per token (a token is a chunk of text, roughly ¾ of a word). Longer conversations and bigger files use more tokens.

Type /usage to see your current usage, your limits, and a breakdown of what's using the most.

Make your usage go further

  • /clear between unrelated tasks. Old conversation history is re-read with every message.
  • /compact during long tasks to shrink the history.
  • Use a lighter model for simple jobs: /model haiku or /model sonnet for routine work, and save the most powerful model for hard problems.
  • Be specific. A precise request avoids wasted exploration.
  • Keep CLAUDE.md lean. It's loaded into every session.

Tip

If you hit a limit mid-task, /rate-limit-options shows your choices: wait for the reset, add usage credits, or upgrade.

Troubleshooting

Problem Try this
claude: command not found Open a new terminal window. Still failing? Re-run the installer and follow the PATH fix it prints, or run claude doctor.
'irm' is not recognized You're in CMD, not PowerShell. Use the CMD install command, or open PowerShell.
The token '&&' is not a valid statement separator You're in PowerShell. Use the PowerShell command instead.
Login window doesn't open Copy the link Claude Code prints into your browser manually.
Claude keeps asking permission for the same thing Choose Yes, and don't ask again, add an allow rule, or run /fewer-permission-prompts.
Claude ignores an instruction Put it in CLAUDE.md, make it concrete, and keep the file short. For must-never-happen rules, use a deny rule.
Responses getting slow or confused /compact or /clear, then restate the goal.
Something is broken and you don't know why /doctor inside Claude Code. It diagnoses and offers fixes.
MCP server shows "Failed to connect" claude mcp get <name> shows the error; re-authenticate in /mcp.

Note

Stuck? Type /bug to report a problem to Anthropic, ask Claude itself ("why did that fail?"), or take the free Claude Code courses at academy.claude.com.

Quick Reference {.unnumbered}

:::cheatsheet Install (Mac/Linux): curl -fsSL https://claude.ai/install.sh | bash
Install (Windows PowerShell): irm https://claude.ai/install.ps1 | iex
Start: cd my-project then claude · Continue: claude -c · Resume: claude -r

Stop: Esc · Undo: Esc Esc or /rewind · Modes: Shift + Tab · New line: Shift + Enter

Plan first: /plan … · Fresh start: /clear · Shrink history: /compact · Project memory: /init

Connect an app: claude mcp add --transport http <name> <url> then /mcp to sign in

Health check: claude doctor · Update: claude update · Usage: /usage :::

The Complete Slash-Command Reference

Slash commands are the control panel of Claude Code. You type / at the start of a message and a searchable menu appears. This chapter lists every built-in command and bundled skill documented by Anthropic as of this edition, so you can look any of them up on paper.

:::summary In this chapter

  • How slash commands, bundled skills, and your own skills differ
  • Deep dives on the commands you'll use every day: /clear, /compact, /context, /rewind, /resume, /model, /init, /memory, /permissions
  • A full A–Z table of every documented command
  • Recipes that chain commands together :::

[!NOTE] Where this list comes from The A–Z table was generated from Anthropic's official commands page, code.claude.com/docs/en/commands, in October 2026. Claude Code updates often. Some commands appear only on certain plans, platforms, or versions. If a command here doesn't show up for you, run claude update, then type / and search for it. Your own / menu is always the final word.

Three kinds of slash command

Kind What it is Example
Built-in command Fixed logic inside Claude Code. It runs instantly and usually doesn't cost tokens. /clear, /model, /exit
Bundled skill A ready-made set of instructions that Claude follows using its tools. It uses tokens, like a normal request. /code-review, /debug, /batch
Your own skill A SKILL.md file you or your team wrote (Chapter 9). /deploy, /fix-issue 123

Anything you type after the command name is passed in as arguments. For example, in /compact keep the API decisions, the text after /compact tells the summary what to focus on.

Tip

A command only runs when it's at the start of your message. If you write "please /deploy this" in the middle of a sentence, nothing runs on its own. Claude just treats it as permission to use that skill if it decides to.

/clear: start fresh

/clear wipes the conversation and starts a new one with empty context. Your files are not touched. Only Claude's short-term memory of the chat is reset. CLAUDE.md loads again automatically.

:::steps Use /clear between tasks

  1. Finish or commit the current task. Ask Claude to summarize what it did if you want a record.
  2. Type /clear. Optionally add a name, such as /clear login-work, to label the old conversation so you can find it later.
  3. Start the next task. Claude now has a clean, fast, cheap context.
  4. Need the old chat back? Use /resume and pick it from the list. :::

Use /clear when you switch to an unrelated task, when Claude keeps repeating an old mistake, or when the conversation has drifted. /reset and /new are aliases.

/compact: shrink without forgetting

Every message, file, and command output takes up room in the context window, the AI's working memory. When it fills up, answers get slower, more expensive, and less sharp. /compact replaces the conversation so far with a summary, so you keep the important facts and free most of the space.

/compact Keep the database schema decisions and the list of failing tests. Drop the exploration of old files.
Situation Use
Same task, but the chat is long /compact
New, unrelated task /clear
Claude went down a wrong path a few steps ago /rewind
You want a side answer without cluttering the chat /btw

Claude Code also auto-compacts when the window is nearly full. /autocompact lets you set how full it gets before that happens. Pass a size such as 500k, or auto to go back to the default.

[!WARNING] Summaries lose detail Compaction is lossy. Exact error messages, file paths, and small decisions can disappear. Before compacting, put anything that must survive into CLAUDE.md or a notes file, or list it in your /compact instructions.

/context: see what's using space

/context draws a coloured grid of your context window, broken down by system prompt, tools, MCP servers, memory files, skills, and messages. It also suggests what to trim. Check it whenever Claude feels slow or forgetful. If MCP tools take up a big block, disconnect the servers you aren't using with /mcp.

/rewind: the undo button

/rewind (aliases /checkpoint and /undo) rolls the conversation, your code, or both back to an earlier message. Pressing Esc twice on an empty prompt opens the same menu.

:::steps Undo a bad change

  1. Type /rewind. A list of your earlier messages appears.
  2. Pick the last good point. Choose the message before things went wrong.
  3. Choose what to restore. Code only, conversation only, or both.
  4. Re-ask, more precisely. Add the detail that was missing the first time. :::

Caution

Checkpoints track edits Claude made with its file tools. Changes made by shell commands, such as rm or a database migration, and edits from background forked skills are not undone. Git is your real safety net, so commit often.

/resume, /branch, /fork, and /rename

Command Use it to…
/resume Reopen an earlier conversation by name or from a picker
/rename my-feature Give this session a memorable name
/branch Split the conversation here so you can try another direction, keeping the original
/fork Copy the conversation into a new background session and keep working here

/model and /effort

/model switches which Claude model you're using and saves it as the default for new sessions. /effort sets how hard the model thinks: low, medium, high, xhigh, max, or auto. Higher effort is slower and uses more of your plan, but it's better for tricky bugs and design work. A good habit is to plan with a strong model at high effort, then do routine edits at a lower setting.

/init and /memory

/init reads your project and writes a starter CLAUDE.md. /memory opens your CLAUDE.md files for editing and lets you turn auto memory on or off and browse what it has saved. Run /init once per project, then keep the file current with /memory.

/permissions, /sandbox, and /fewer-permission-prompts

/permissions shows and edits your allow, ask, and deny rules (Chapter 6). /sandbox toggles sandbox mode on supported systems. The bundled /fewer-permission-prompts skill scans your history for safe, read-only commands you approve often, then suggests allow rules for them. Review that list before you accept it.

Review and quality commands

Command What it's for
/diff See every change in your working tree
/code-review (alias /review) Have Claude review your diff, a branch, or a PR for bugs
/security-review Scan your branch's changes for security risks
/simplify Run parallel reviewers that tidy up changed code
/verify and /run Build and run your app to confirm a change actually works

Housekeeping commands

/doctor runs a setup checkup and offers fixes. /usage (aliases /cost and /stats) shows plan limits and spending. /status shows your version, model, and account. /config opens settings. /export saves the chat as text, /copy copies the last answer, and /release-notes shows what's new. /exit (or /quit) leaves.

Recipes: commands that work well together

:::cheatsheet Long feature, keep it sharp: /plan → approve → work → /compact keep decisions → continue → /code-review → commit.

Switching tasks: commit → /clear new-task → /context to confirm it's empty.

Undo a wrong turn: /rewind → restore code and conversation → re-prompt with more detail.

Before a pull request: /diff → /code-review → /security-review → /verify.

Something's broken in Claude Code itself: /doctor → /debug describe the problem → /bug if it's a real bug.

Running low on plan usage: /usage → /model to a lighter model → /effort low for routine edits. :::

Every documented command, A–Z

The Kind column shows whether the command is built in, a bundled skill (uses tokens as Claude works), or a workflow. Square brackets [ ] mean the argument is optional. Angle brackets < > mean it's required.

:::compact

Command Kind What it does
/add-dir <path> Built-in Add a working directory for file access during the current session.
/advisor [model|off] Built-in Enable or disable the advisor tool, which consults a second model for guidance at key moments during a task.
/agents Built-in Print a reminder to ask Claude to create or manage subagents, or to edit .claude/agents/ or ~/.claude/agents/ directly.
/artifact-capabilities Skill Load the reference for the runtime capabilities a published artifact can use, such as calling your connectors or offering a file download, including which ones your account has.
/artifact-diagramming Skill Load diagramming guidance for Claude to follow in artifacts: when a diagram helps, what to draw, and how to write inline SVG that stays legible in light and dark themes.
/artifacts Built-in List the artifacts you own or that are shared with you, then attach one to the session, open it in your browser, or copy its link.
/auto-mode-setup Built-in Draft autoMode.environment entries from your project and recent sessions, then review the draft and save it to your user settings.
/autocompact [auto|<tokens>] Built-in Set the auto-compact window: how full the context window gets before Claude Code compacts automatically.
/autofix-pr [prompt] Built-in Spawn a cloud session that watches the current branch's PR and pushes fixes when CI fails or reviewers leave comments.
/background [prompt] Built-in Detach the current session to run as a background agent and free this terminal. Alias: /bg.
/batch <instruction> Skill Orchestrate large-scale changes across a codebase in parallel.
/branch [name] Built-in Create a branch of the current conversation at this point, so you can try a different direction without losing the conversation as it stands.
/btw [question] Built-in Ask a side question about the current session without adding to the conversation.
/bug [report] Built-in Report a bug or share your conversation. Alias: /share.
/cd <path> Built-in Move this session to a new working directory, keeping the conversation.
/chrome Built-in Configure Claude in Chrome settings.
/claude-api [migrate|upgrade|managed-agents-onboard|prompt-audit|cost-optimize|build-eval|hillclimb|preserved-thinking-migration] Skill Load Claude API and Managed Agents reference material for your project's language.
/claude-in-chrome [task] Skill Have Claude carry out a task in your browser, such as testing a page, filling a form, or reading console logs, through Claude in Chrome.
/clear [name] Built-in Start a new conversation with empty context. Alias: /reset, /new.
/code-review [low|medium|high|xhigh|max|ultra] [--fix] [--comment] [--max-findings n|all|default] [pr#|branch|path] Skill Review the current diff, or a PR number, branch, or path you pass, for correctness bugs. Alias: /review.
/color [color|default] Built-in Set the prompt bar color for the current session.
/compact [instructions] Built-in Free up context by summarizing the conversation so far.
/config [key=value ...] Built-in Open the Settings interface to adjust theme, model, output style, and other preferences. Alias: /settings.
/context [all] Built-in Visualize current context usage as a colored grid.
/copy [N] Built-in Copy the last assistant response to clipboard.
/cost Built-in Alias for /usage.
/dataviz [request] Skill Design guidance for charts, graphs, and dashboards.
/debug [description] Skill Enable debug logging for the current session and troubleshoot issues by reading the session debug log.
/deep-research <question> Workflow Fan out web searches on a question, fetch and cross-check sources, and synthesize a cited report.
/design [brief] Skill Draft UI mockups, screen flows, landing pages, or posters as artboards on one canvas, published as a Claude Design artifact, for example /design a settings screen for a mobile banking app.
/design-login Built-in Authorize design-system access for /design-sync with your claude.ai account.
/design-sync [hint] Skill Convert your repo's React design system and upload it to Claude Design, so designs it produces use your real components.
/desktop Built-in Continue the current session in the Claude Code Desktop app. Alias: /app.
/diff Built-in Review the changes in your working tree, including the edits Claude has made so far.
/doctor [prompt-audit [path]] Skill Run a setup checkup that diagnoses installation, settings, extension, and CLAUDE.md problems and proposes fixes that Claude applies after you confirm. Alias: /checkup.
/effort [level|auto|status|ultracode [on|off]] Built-in Set the effort level: low to xhigh, max, or auto; status prints it. ultracode or ultracode on turns ultracode on for the session at the current level, and ultracode off turns it off; the ultracode key…
/exit Built-in Exit the CLI. In an attached background session, this detaches and the session keeps running. Alias: /quit.
/export [filename] Built-in Export the current conversation as plain text.
/fast [on|off] Built-in Toggle fast mode on or off.
/feedback [report] Built-in Send product feedback about Claude Code.
/fewer-permission-prompts Skill Scan your transcripts for common read-only Bash and MCP tool calls, then add a prioritized allowlist to project .claude/settings.json to reduce permission prompts.
/focus Built-in Toggle the focus view, which shows only your last prompt, a one-line tool-call summary with edit diffstats, and the final response.
/fork [prompt] Built-in Copy the current conversation into a new background session and keep working here.
/goal [condition|clear] Built-in Set a goal: Claude keeps working across turns until the condition is met or the goal clears for another reason.
/heapdump Built-in Write a JavaScript heap snapshot and a memory breakdown to ~/Desktop, or your home directory on Linux without a Desktop folder, for diagnosing high memory usage.
/help Built-in Show help and available commands.
/hooks Built-in View hook configurations.
/ide Built-in Manage IDE integrations and show status.
/import [codex|gemini|cursor] [--dry-run] [--yes] Built-in Bring configuration from OpenAI Codex, Google Gemini CLI, or Cursor on your machine into Claude Code, including instruction files, MCP servers, commands, subagents, and skills.
/init Built-in Initialize project with a CLAUDE.md guide.
/insights Built-in Generate an HTML report analyzing your recent sessions on this machine: which projects you work in, how you use Claude Code, where things go wrong, and features to try.
/install-github-app Built-in Install the Claude GitHub App for a repository, with an optional step to set up GitHub Actions workflows and secrets.
/install-slack-app Built-in Install the Claude Slack app.
/keybindings Built-in Open your keyboard shortcuts file.
/list-agents Built-in List the subagents, agent team teammates, and other Claude Code sessions Claude can message, with the name to use for each.
/login Built-in Sign in to your Anthropic account.
/logout Built-in Sign out from your Anthropic account.
/loop [interval] [prompt] Skill Run a prompt repeatedly while the session stays open. Alias: /proactive.
/mcp [reconnect (<server>|all)|enable|disable [<server>|all]] Built-in Manage MCP server connections and OAuth authentication.
/memory Built-in Edit CLAUDE.md files, enable or disable auto memory, and view auto memory entries.
/mobile Built-in Show QR code to download the Claude mobile app. Alias: /ios, /android.
/model [model] Built-in Switch the AI model and save it as your default for new sessions.
/output-style [style] Built-in List output styles or switch to one, for example /output-style concise.
/passes Built-in Share a free week of Claude Code with friends.
/permissions Built-in Manage allow, ask, and deny rules for tool permissions. Alias: /allowed-tools.
/plan [description] Built-in Enter plan mode directly from the prompt.
/plugin [subcommand] Built-in Manage Claude Code plugins.
/plugin-authoring Built-in Load the reference Claude works from to write a mod.
/powerup Built-in Discover Claude Code features through quick interactive lessons with animated demos.
/pr-comments [PR] Built-in Removed in v2.1.91. Ask Claude directly to view pull request comments instead.
/privacy-settings Built-in View and update your privacy settings.
/radio Built-in Open Claude FM lo-fi radio in your browser.
/rate-limit-options Built-in Show ways to keep working when a claude.ai usage limit blocks a request: wait and continue automatically when the limit resets, add usage credits, or upgrade your plan.
/recap Built-in Generate a one-line summary of the current session on demand.
/release-notes Built-in View the changelog in an interactive version picker.
/reload-plugins [--force] Built-in Reload all active plugins to apply pending changes without restarting.
/reload-skills Built-in Re-scan skill and command directories so skills added or changed on disk during the session become available without restarting.
/remote-control Built-in Make this session available for Remote Control from claude.ai. Alias: /rc.
/remote-env Built-in Choose the default cloud environment for cloud sessions you start from the CLI.
/rename [name] Built-in Rename the current session and show the name on the prompt bar.
/resume [session] Built-in Resume a conversation by ID or name, or open the session picker. Alias: /continue.
/review [low|medium|high|xhigh|max|ultra] [--fix] [--comment] [--max-findings n|all|default] [pr#|branch|path] Built-in Alias of /code-review: reviews the current diff, or a PR number, branch, or path you pass, such as /review 1234, and takes the same effort levels and flags.
/rewind Built-in Rewind the conversation and/or code to a previous point, or summarize from a selected message. Alias: /checkpoint, /undo.
/run Skill Launch and drive your project's app to see a change working, not only passing tests.
/run-skill-generator Skill Teach /run and /verify how to build, launch, and drive your project's app from a clean environment by writing a per-project skill.
/sandbox Built-in Toggle sandbox mode.
/schedule [description] Built-in Create, update, list, or run routines, which execute in the cloud. Alias: /routines.
/scroll-speed Built-in Adjust mouse wheel scroll speed interactively.
/security-review Built-in Analyze the changes on your current branch for security vulnerabilities.
/setup-bedrock Built-in Configure Amazon Bedrock authentication, region, and model pins through an interactive wizard.
/setup-vertex Built-in Configure Google Cloud's Agent Platform authentication, project, region, and model pins through an interactive wizard.
/simplify [target] Skill Review the changed code for cleanup opportunities and apply the fixes.
/skill-doctor Built-in Show what each of your skills costs in context and how often it gets used, so you can find skills to turn off.
/skills Built-in List available skills.
/slides [brief] Skill Make a new presentation as a Claude Slides artifact filled from your brief, for example /slides a quarterly review of the platform team.
/stats Built-in Alias for /usage.
/status Built-in Open the Settings interface on the Status tab, showing version, model, account, and connectivity.
/statusline Built-in Configure Claude Code's status line.
/stickers Built-in Order Claude Code stickers.
/stop Built-in Stop the background session you're attached to, or the one you send it to as a peek reply; the transcript and any worktree are kept.
/subtask <task> Built-in Spawn a forked subagent: a background subagent that inherits the full conversation and works on the task while you keep working.
/tasks Built-in View and manage background work in the current session, including subagents that have finished.
/team-onboarding Built-in Generate a team onboarding guide from your Claude Code usage history.
/teleport Built-in Pull a cloud session into this terminal.
/terminal-setup Built-in Install a Shift+Enter keybinding for newlines in VS Code, Cursor, Devin Desktop, Alacritty, or Zed.
/theme Built-in Change the color theme.
/tui [default|fullscreen] Built-in Set the terminal UI renderer and relaunch into it with your conversation intact. fullscreen enables the flicker-free alt-screen renderer.
/ultraplan <prompt> Built-in Removed.
/ultrareview [PR or branch] Built-in Run a deep, multi-agent code review in a cloud sandbox with ultrareview.
/update-config [request] Skill Describe a settings change, such as allowing a command, setting an environment variable, or adding a hook, and Claude edits the matching settings.json file.
/upgrade Built-in Open the upgrade page in your browser to switch to a higher plan tier.
/usage Built-in Show session cost, plan usage limits, and activity stats.
/usage-credits Built-in Configure usage credits, or request them from your admin, when you hit a limit.
/verify Skill Confirm a code change does what it should by building your project's app, running it, and observing the result, rather than relying on tests or type checks.
/vim Built-in Removed in v2.1.92. To toggle between Vim and Normal editing modes, use /config → Editor mode.
/voice [hold|tap|off] Built-in Toggle voice dictation, or enable it in a specific mode.
/web-setup Built-in Connect your GitHub account for cloud sessions using your local gh CLI credentials.
/workflow-authoring Skill Load the reference for writing dynamic workflow scripts: the script API, resume behavior, quality patterns, and worked examples.
/workflows Built-in Open the workflow progress view to watch, pause, resume, or save running and completed workflows.
:::

Note

A few rows say "Removed." They're kept here because older tutorials still mention them. For example, /vim was replaced by Editor mode in /config.

Command-Line Flags and Shortcuts Reference

This chapter is the companion to the slash-command reference. It covers what you can type in your terminal before Claude Code starts, and every keyboard shortcut you can press inside it. The tables were generated from Anthropic's CLI reference (code.claude.com/docs/en/cli-reference) and interactive-mode page (code.claude.com/docs/en/interactive-mode) in October 2026.

:::summary In this chapter

  • Every claude … subcommand
  • Every startup flag, with the ones beginners need marked below
  • Every keyboard shortcut and line-editing key :::

The ten flags beginners actually use

Flag Why you'd use it
-c / --continue Pick up exactly where you left off in this folder
-r / --resume Choose an older conversation
-p / --print Ask one question, print the answer, exit, which is great for scripts
--model Pick the model for this session only
--permission-mode plan Start in plan mode, so nothing gets edited until you approve
--add-dir ../other Let Claude also work in a second folder
--effort high Ask for deeper thinking this session
--debug Write a detailed log when something misbehaves
--chrome Turn on browser control for web testing
--dangerously-skip-permissions Skip all prompts. Only use it in a throwaway container or VM (see Vol. 8)

Caution

Flags with "dangerously" in their name remove safety checks. On your everyday computer, use plan mode, accept-edits mode, or allow rules instead.

Terminal subcommands

:::compact

Command What it does
claude Start interactive session.
claude "query" Start interactive session with initial prompt.
claude -p "query" Query via SDK, then exit.
claude -c Continue most recent conversation in current directory.
claude -c -p "query" Continue via SDK.
claude -r "<session>" "query" Resume session by ID or name.
claude update Update to latest version.
claude gateway Start the self-hosted Claude apps gateway server, for administrators deploying SSO and policy in front of Claude Code on Amazon Bedrock, Google Cloud's Agent Platform, or Microsoft Foundry.
claude install [version] Install or reinstall the native binary.
claude auth login Sign in to your Anthropic account.
claude auth logout Log out from your Anthropic account.
claude auth status Show authentication status as JSON. Use --text for human-readable output.
claude agents Open agent view to monitor and dispatch parallel background sessions.
claude attach <id|name> Attach to a background session in this terminal.
claude auto-mode defaults Print the built-in auto mode classifier rules as JSON. Use claude auto-mode config to see your effective config with settings applied. --label <prefix> prints only the rules whose label starts…
claude auto-mode reset Restore the default auto mode configuration by removing the autoMode section from your user settings file.
claude daemon logs Follow the background-session supervisor's log file, ~/.claude/daemon.log, printing new lines as they arrive until you press Ctrl+C.
claude daemon run Run the background-session supervisor in the foreground of this terminal, printing its log.
claude daemon status Print the background-session supervisor's state, version, socket directory, and worker count for diagnostics.
claude daemon stop --any Stop the background-session supervisor and the sessions it hosts.
claude doctor Print read-only installation and settings diagnostics from the terminal without starting a session, including install health, settings-file validation errors, and Remote Control eligibility.
claude import [source] Start an interactive session that runs /import to bring configuration from other coding agents into Claude Code.
claude logs <id|name> Print recent output from a background session.
claude mcp Configure Model Context Protocol (MCP) servers.
claude mcp login <name> Run a configured MCP server's OAuth flow without opening the interactive /mcp panel.
claude mcp logout <name> Clear stored OAuth credentials for an MCP server.
claude plugin Manage Claude Code plugins.
claude purge [path] Delete all local Claude Code state for a project: transcripts, task lists, debug logs, file-edit history, prompt history lines, and the project's entry in ~/.claude.json.
claude remote-control Start a Remote Control server to control Claude Code from Claude.ai or the Claude app.
claude respawn <id> Restart a background session, running or stopped, with its conversation intact.
claude rm <id> Remove a background session from the list.
claude self-hosted-runner Start a runner process that registers this machine or container with a self-hosted environment and hosts Claude Code cloud sessions on your infrastructure.
claude setup-token Generate a long-lived OAuth token for CI and scripts.
claude stop <id> Stop a background session.
claude ultrareview [target] Run ultrareview non-interactively.
:::

Every startup flag

:::compact

Flag What it does
--add-dir Add additional working directories for Claude to read and edit files.
--advisor <model> Enable the server-side advisor tool for this session with a model alias, fable, opus, or sonnet, or a full model ID. Takes precedence over the advisorModel…
--agent Specify an agent for the current session (overrides the agent setting).
--agents Define custom subagents dynamically via JSON. Accepts the fields listed for CLI-defined subagents.
--allow-dangerously-skip-permissions Add bypassPermissions to the Shift+Tab mode cycle without starting in it.
--allowedTools, --allowed-tools Tools that execute without prompting for permission, apart from reads from network paths.
--append-subagent-system-prompt Append custom text to the end of every subagent's system prompt, nested subagents included, apart from a forked subagent, which reuses the conversation's own prompt.
--append-subagent-system-prompt-file Load text from a file and append it to subagent system prompts.
--append-system-prompt Append custom text to the end of the default system prompt.
--append-system-prompt-file Load additional system prompt text from a file and append to the default prompt.
--autocompact <auto|tokens> Set the auto-compact window for this session without changing your saved settings.
--ax-screen-reader Render screen-reader friendly output: flat text without decorative borders or animations.
--bare Minimal mode: skip auto-discovery of hooks, skills, custom commands, subagents, installed plugins, MCP servers, auto memory, and CLAUDE.md so scripted calls start faster.
--betas Beta headers to include in API requests (API key users only).
--bg, --background Start the session as a background agent and return immediately.
--channels (Research preview) MCP servers whose channel notifications Claude should listen for in this session.
--chrome Enable Chrome browser integration for web automation and testing.
--cloud With a task description, create a new cloud session.
--continue, -c Load the most recent conversation in the current directory, including a background session that has finished; opening finished background sessions requires Claude…
--dangerously-load-development-channels Enable channels that are not on the approved allowlist, for local development.
--dangerously-skip-permissions Skip permission prompts.
--debug Enable debug mode with optional category filtering, such as --debug='mcp,startup' or --debug='!1p'.
--debug-file <path> Write debug logs to a specific file path.
--desktop Open the Claude Desktop app on the current directory and exit without starting a session in the terminal.
--disable-slash-commands Disable all skills and commands for this session.
--disallowedTools, --disallowed-tools Deny rules.
--effort Set the effort level for the current session.
--enable-auto-mode Removed in v2.1.111. Auto mode is now in the Shift+Tab cycle by default; use --permission-mode auto to start in it.
--environment <environment-id> Create a new cloud session that runs on the self-hosted environment with the given ID. Environment IDs start with ccpool_.
--exclude-dynamic-system-prompt-sections Move per-user context, such as the auto memory location, out of the system prompt and into the first user message.
--exec Run a shell command as a PTY-backed background job instead of starting a Claude session.
--fallback-model Enable automatic fallback to the specified model(s) when the primary model is overloaded or not available, for example a retired model.
--fork-session When resuming, create a new session ID instead of reusing the original (use with --resume or --continue).
--forward-subagent-text Emit subagent text and thinking blocks in the output stream as assistant and user messages with parent_tool_use_id set, so you can reconstruct each subagent's…
--from-pr Open the session picker filtered to sessions linked to a specific pull request.
--ide Automatically connect to IDE on startup if exactly one valid IDE is available.
--init Run Setup hooks with the init matcher before the session (print mode only).
--init-only Run Setup and SessionStart hooks, then exit without starting a conversation.
--include-hook-events Include hook lifecycle events in the output stream. SessionStart and Setup hook events are always included and don't need this flag.
--include-partial-messages Include partial streaming events in output.
--input-format Specify input format for print mode (options: text, stream-json).
--json-schema Get validated JSON output matching a JSON Schema after the agent completes its workflow (print mode only).
--maintenance Run Setup hooks with the maintenance matcher before the session (print mode only).
--max-budget-usd Stop the run once estimated spend on API calls reaches this amount (print mode only).
--max-turns Limit the number of agentic turns (print mode only).
--mcp-config Load MCP servers from JSON files or strings (space-separated).
--model Sets the model for the current session with a model alias such as sonnet, opus, haiku, or fable, or a model's full name.
--name, -n Set a display name for the session, shown in /resume and the terminal title.
--no-chrome Disable Chrome browser integration for this session.
--no-session-persistence Disable session persistence so sessions are not saved to disk and cannot be resumed.
--output-format Specify output format for print mode (options: text, json, stream-json).
--permission-mode Begin in a specified permission mode.
--permission-prompt-tool Specify an MCP tool to handle permission prompts in non-interactive mode.
--permission-prompts Set who answers permission prompts in print mode.
--plugin-dir Load a plugin from a directory or .zip archive, or several from a folder of plugins, for this session only.
--plugin-url Fetch a plugin .zip archive from a URL for this session only.
--print, -p Print response without interactive mode (see Agent SDK documentation for programmatic usage details).
--prompt-suggestions Emit a prompt_suggestion message with a predicted next user prompt after each turn that generates one; very short conversations can produce none.
--ref <branch> With --environment, base the new session's checkout on a named ref instead of local HEAD.
--remote Deprecated alias for --cloud, including the existing-session form.
--remote-control, --rc Start an interactive session with Remote Control enabled so you can also control it from claude.ai or the Claude app.
--remote-control-session-name-prefix <prefix> Prefix for auto-generated Remote Control session names when no explicit name is set.
--replay-user-messages Re-emit user messages from stdin back on stdout for acknowledgment.
--restricted Start in restricted mode.
--resume, -r Resume a specific session by ID or name, or show an interactive picker to choose a session.
--safe-mode Start with all customizations disabled to troubleshoot a broken configuration: CLAUDE.md, skills, plugins, hooks, MCP servers, custom commands and agents, output…
--session-id Use a specific session ID for the conversation (must be a valid UUID).
--setting-sources Comma-separated list of setting sources to load (user, project, local).
--settings Path to a settings JSON file or an inline JSON string.
--strict-mcp-config Only use MCP servers from --mcp-config, ignoring all other MCP configurations.
--system-prompt Replace the entire system prompt with custom text.
--system-prompt-file Load system prompt from a file, replacing the default prompt.
--system-prompt-snapshot Pass off to rebuild the system prompt on every request instead of reusing the prompt recorded on the conversation's first request, for example while you iterate on…
--teleport Resume a cloud session in your local terminal.
--teammate-mode Set how agent team teammates display: in-process (default), auto, tmux, or iterm2.
--tmux Create a tmux session for the worktree.
--tools Restrict which built-in tools Claude can use.
--verbose Enable verbose logging, shows full turn-by-turn output.
--version, -v Output the version number.
--worktree, -w Start Claude in an isolated git worktree at <repo>/.claude/worktrees/<name>.
--system-prompt Replaces the entire default prompt.
--system-prompt-file Replaces with file contents.
--append-system-prompt Appends to the default prompt.
--append-system-prompt-file Appends file contents to the default prompt.
--system-prompt-snapshot With off, rebuilds the prompt on every request.
:::

Every keyboard shortcut

:::compact

Keys What it does
Ctrl+C Interrupt, or clear input.
Ctrl+X Ctrl+K Stop all running background subagents in this session, and turn off artifact auto-replies for the rest of it.
Ctrl+D Exit Claude Code session.
Ctrl+G or Ctrl+X Ctrl+E Open in default text editor.
Ctrl+L Redraw the screen.
Ctrl+O Toggle transcript viewer.
Ctrl+R Reverse search command history.
Ctrl+V or Cmd+V (iTerm2) or Alt+V (Windows and WSL) Paste image from clipboard.
Ctrl+B Background running tasks.
Ctrl+T Toggle Claude's task checklist.
Ctrl+S Stash or restore prompt.
Ctrl+Z Suspend Claude Code.
Left/Right arrows Cycle through dialog tabs.
Tab Accept an autocomplete suggestion, or add a comment to a permission answer.
Up/Down arrows or Ctrl+P/Ctrl+N Move cursor or navigate command history.
Esc Interrupt Claude, or close a dialog.
Esc + Esc Clear input draft, or rewind.
Ctrl+Enter or Ctrl+X Ctrl+S Send queued messages now.
Shift+Tab, or Alt+M on Windows when the Node or Bun runtime doesn't enable VT input mode Cycle permission modes.
Option+P (macOS) or Alt+P (Windows/Linux) Switch model.
Option+T (macOS) or Alt+T (Windows/Linux) Toggle extended thinking.
Option+O (macOS) or Alt+O (Windows/Linux) Toggle fast mode.
Ctrl+A Move cursor to start of current line.
Ctrl+E Move cursor to end of current line.
Ctrl+K Delete to end of line.
Ctrl+U Delete from cursor to line start.
Ctrl+W Delete back to previous whitespace.
Ctrl+Y Paste deleted text.
Alt+Y (after Ctrl+Y) Cycle paste history.
Alt+B Move cursor back one word.
Alt+F Move cursor forward one word.
Alt+D Delete to end of word.
Ctrl+_ or Ctrl+Shift+- Undo last input edit.
Ctrl+T Toggle syntax highlighting for code blocks.
/ at start Command or skill.
! at start Shell mode.
@ File path mention.
: Emoji shortcode.
? on empty input Toggle the shortcut help panel.
? Toggle the keyboard shortcut help panel.
{ / } Jump to the previous or next user prompt, like vim paragraph motion.
Ctrl+E Toggle show all content.
[ Write the full conversation to your terminal's native scrollback so Cmd+F, tmux copy mode, and other native tools…
v Write the conversation to a temporary file and open it in $VISUAL or $EDITOR.
q, Ctrl+C, Esc Exit transcript view.
Esc or Ctrl+[ Enter NORMAL mode.
i Insert before cursor.
I Insert at beginning of line.
a Insert after cursor.
A Insert at end of line.
o Open line below.
O Open line above.
v Start character-wise visual selection.
V Start line-wise visual selection.
h/j/k/l Move left/down/up/right.
Space Move right.
w Next word.
e End of word.
b Previous word.
0 Beginning of line.
$ End of line.
^ First non-blank character.
gg Beginning of input.
G Beginning of last line.
f{char} Jump to next occurrence of character on the current line.
F{char} Jump to previous occurrence of character on the current line.
t{char} Jump to just before next occurrence of character on the current line.
T{char} Jump to just after previous occurrence of character on the current line.
; Repeat last f/F/t/T motion.
, Repeat last f/F/t/T motion in reverse.
% Jump to the matching (), [], or {} bracket, starting from the next bracket on the current line when the cursor…
/ Open reverse history search, same as Ctrl+R.
x Delete character.
r{char} Replace character under cursor with {char}.
dd Delete line.
D Delete to end of line.
dw/de/db Delete word/to end/back.
df{char}/dt{char} Delete to and including, or up to, the next occurrence of a character on the current line.
dj/dk Delete the current line and the line below or above.
dgg/dG Delete from the current line to the first or last line.
d0/c0/y0 Delete, change, or yank from the cursor back to the beginning of the line.
d%/c%/y% Delete, change, or yank from the cursor through the bracket that % jumps to.
cc Change line.
C Change to end of line.
cw/ce/cb Change word/to end/back.
s Substitute character: delete the character under the cursor and enter INSERT mode.
S Substitute line: clear the line and enter INSERT mode.
yy/Y Yank (copy) line.
yw/ye/yb Yank word/to end/back.
p Paste after cursor.
P Paste before cursor.
>> Indent line.
<< Dedent line.
J Join lines.
u Undo.
. Repeat last change.
iw/aw Inner/around word.
iW/aW Inner/around WORD (whitespace-delimited).
i"/a" Inner/around double quotes.
i'/a' Inner/around single quotes.
i(/a( Inner/around parentheses.
i[/a[ Inner/around brackets.
i{/a{ Inner/around braces.
d/x Delete selection.
y Yank selection.
c/s Change selection.
p Replace selection with register contents.
r{char} Replace every selected character with {char}.
~/u/U Toggle, lowercase, or uppercase selection.
>/< Indent or dedent selected lines.
J Join selected lines.
o Swap cursor and anchor.
iw/aw/i"/… Select a text object.
v/V Toggle between character-wise and line-wise, or exit.
Space, Enter, Escape Dismiss the answer and return to the prompt.
Up / Down Scroll the answer.
Shift+Left / Shift+Right Step between this answer and your earlier /btw answers. Shift+Left moves to older answers and Shift+Right…
c Copy the answer to your clipboard as raw Markdown.
f Start a forked subagent that inherits the parent conversation plus this question and answer, so it can continue with…
x Clear the list of earlier /btw exchanges shown above the current answer.
:::

Real-World Workflows, Step by Step

Knowing the commands is half the job. This chapter strings them into complete, repeatable workflows you can follow with the book open next to your keyboard.

:::summary In this chapter

  • Fix a bug from an error message
  • Add a feature safely with plan mode
  • Understand a project you've never seen
  • Write and run tests
  • Ship a pull request
  • Run long tasks in the background
  • Automate with claude -p :::

Workflow 1: Fix a bug from an error message

:::steps

  1. Reproduce it. Run the program and copy the full error, including the lines above it.
  2. Start Claude in the project folder. Type claude.
  3. Paste the error with context.
    When I run `npm start` I get the error below. Find the cause, explain it in plain English, then propose the smallest fix. Don't change anything else.
    [paste error]
    
  4. Read the explanation before approving. If it doesn't make sense to you, ask "why?" until it does.
  5. Approve the edit, then ask Claude to rerun the command. A fix isn't done until the error is gone.
  6. Commit. Ask: "Commit this with a clear message." :::

Workflow 2: Add a feature with plan mode

:::steps

  1. Enter plan mode. Press Shift + Tab until the footer says plan, or type /plan add a dark-mode toggle.
  2. Let Claude research. It reads files but can't edit them.
  3. Review the plan. Push back on anything too big. Press Ctrl + G to edit the plan in your text editor.
  4. Approve. Claude switches out of plan mode and starts working.
  5. Check in at milestones. Ask for a short status after each step.
  6. Review with /diff, then /code-review.
  7. Compact or clear. If the session is long and you're continuing, use /compact. If you're done, commit and /clear. :::

Workflow 3: Learn an unfamiliar project

Give me a beginner-friendly tour of this project: what it does, how to run it, the main folders, and the five files I should read first. Don't edit anything.

Follow up with "trace what happens when a user clicks Sign up." This trace question is the fastest way to learn real code. When you're done, run /init so the next session starts with what you learned.

Workflow 4: Tests first

:::steps

  1. Ask for tests that describe the behaviour you want, and tell Claude they should fail for now.
  2. Run them and confirm they fail. This proves the tests actually check something.
  3. Ask Claude to make the tests pass without changing the tests.
  4. Run the full test suite so you know nothing else broke. :::

Tip

Saying "don't modify the tests" stops a common shortcut, where an AI "fixes" a failing test by weakening it.

Workflow 5: Ship a pull request

:::steps

  1. Branch first. "Create a branch called feature/contact-form."
  2. Do the work (Workflow 2).
  3. Review. /diff → /code-review → /security-review.
  4. Commit and push. "Commit in logical chunks and push the branch."
  5. Open the PR. "Open a pull request with a summary and testing notes." This uses the gh tool (Vol. 2).
  6. Address feedback. Paste reviewer comments back to Claude, or try /autofix-pr if your plan supports cloud sessions. :::

Workflow 6: Long tasks in the background

Press Ctrl + B to send a long-running command, like a dev server or test run, to the background. Use /tasks to check on it. To detach the whole session and get your terminal back, use /background. Then watch it from your shell with claude agents. /subtask hands a side job to a helper that sees the full conversation.

Workflow 7: Scripts with claude -p

-p (print mode) runs one request with no interactive screen. That makes it ideal for scripts and scheduled jobs.

claude -p "Summarize what changed in git today in five bullets" > today.md
git diff | claude -p "Write a commit message for this diff"

Warning

In print mode nobody is around to approve prompts. Give it only the permissions it needs, with --allowedTools or allow rules, and never broad shell access on a machine you care about.

Daily routine checklist

:::checklist

  • git status is clean before you start
  • Start with claude -c to continue, or claude for a fresh task
  • One task per conversation
  • Use plan mode for anything touching more than two files
  • /context when things feel slow, then /compact or /clear
  • /diff before every commit
  • Commit small and often
  • End by asking for a summary of what changed and what's next :::

Troubleshooting and FAQ

When something goes wrong, work through this chapter from top to bottom. Most problems are one of a dozen common causes.

:::summary In this chapter

  • A first-aid routine for any problem
  • Install and sign-in fixes
  • When Claude gets slow, forgetful, or stuck
  • Permission and tool problems
  • Frequently asked questions :::

First-aid routine

:::steps

  1. Read the whole message. The fix is often in the last line.
  2. Run claude doctor in your terminal, or /doctor inside a session.
  3. Update. Run claude update, then restart.
  4. Check sign-in. Run claude auth status.
  5. Turn on logging. Restart with claude --debug, or run /debug describe the problem.
  6. Search the official docs at code.claude.com/docs.
  7. Report it. If it's a real bug, use /bug. It shows you what will be sent and asks you to confirm first. :::

Symptom table

:::compact

Symptom Likely cause Fix
command not found: claude The install folder isn't on your PATH Open a new terminal. Run claude doctor. Reinstall with the native installer
Browser sign-in never finishes Pop-up blocked or wrong browser profile Copy the sign-in URL into the right browser. Try claude auth login
"Usage limit reached" Your plan's limit for this period is used up /usage to see when it resets. /rate-limit-options lists choices. Switch to a lighter /model
Answers get vague or repetitive The context window is full of old material /context, then /compact or /clear
Claude keeps redoing a mistake The bad attempt is still in context /rewind to before the mistake. Re-prompt with the missing detail
Too many permission prompts No allow rules yet /permissions to add rules, or /fewer-permission-prompts
Claude edits files you didn't want touched The request was too vague Use plan mode. Add "only change X" to the prompt. Add deny rules
MCP server shows "failed" Bad config or expired sign-in /mcp to reconnect or sign in. Check the command and path
A skill doesn't appear in / Wrong folder, bad YAML, or a name clash Check .claude/skills/<name>/SKILL.md. Run /reload-skills
Hooks don't fire Wrong event name or matcher Run /hooks to inspect them, then --debug
Garbled screen The terminal display got out of sync Ctrl + L to redraw
Pasting an image does nothing Your terminal uses a different key Try Ctrl + V, Cmd + V (iTerm2), or Alt + V (Windows)
Shift + Enter sends instead of adding a new line Your terminal doesn't support it Run /terminal-setup, or use \ + Enter or Ctrl + J
:::

Frequently asked questions

Does /clear delete my files? No. It only clears the conversation. Your files and git history stay as they are.

Is /compact the same as /clear? No. /compact keeps a summary and stays on the same task. /clear starts from nothing.

Can Claude see my whole computer? By default it works in the folder you started it in, plus any folders you add with /add-dir. It asks before it does anything outside its allowed permissions.

Does it remember me between sessions? Only through CLAUDE.md, auto memory, and the conversations you resume. A fresh session doesn't automatically know what you discussed yesterday.

What does it cost? It depends on your plan and changes over time. See the Costs chapter, and run /usage for your own numbers. This manual doesn't quote prices, because they change.

Is my code used for training? That depends on your account type and settings. Check Anthropic's current privacy settings for your plan. Don't take a third-party guide's word for it.

Can I use it on my phone? You can use Claude Code on the web and in the Claude app, and /remote-control lets you drive a session running on your computer. See the "Beyond the Terminal" chapter.

Hooks, Settings, and Environment Reference

This is the deep reference for people who want to automate Claude Code. You don't need it on day one. The tables were generated from Anthropic's hooks reference (code.claude.com/docs/en/hooks) and environment-variable reference (code.claude.com/docs/en/env-vars) in October 2026.

:::summary In this chapter

  • Every hook event and when it fires
  • A worked hook example
  • Where settings files live and which one wins
  • Every documented environment variable :::

Settings files and precedence

File Scope Commit to git?
Managed settings (set by IT) Whole organisation, can't be overridden Not applicable
.claude/settings.local.json You, in this project No, keep it private
.claude/settings.json Everyone on this project Yes
~/.claude/settings.json You, in every project Not applicable

When two files disagree, the more specific one wins: managed, then command-line flags, then local project, then shared project, then user. Deny rules beat allow rules wherever they come from.

Hook events

A hook is a command that runs automatically when an event fires. Use /hooks to see the hooks you have, or /update-config to ask Claude to add one.

:::compact

Event Fires…
SessionStart When a session begins or resumes.
Setup When you start Claude Code with --init-only, or with --init or --maintenance in -p mode.
UserPromptSubmit When a prompt is submitted, before Claude processes it.
UserPromptExpansion When a user-typed command expands into a prompt, before it reaches Claude.
PreToolUse Before a tool call executes.
PermissionRequest When a tool call needs a permission decision.
PermissionDenied When auto mode denies a tool call, including denials without a classifier verdict.
PostToolUse After a tool call succeeds.
PostToolUseFailure After a tool call fails.
PostToolBatch After a full batch of parallel tool calls resolves, before the next model call.
Notification When Claude Code sends a notification.
MessageDisplay While assistant message text is displayed.
SubagentStart When a subagent is spawned.
SubagentStop When a subagent finishes.
TaskCreated When a task is being created via TaskCreate.
TaskCompleted When a task is being marked as completed.
Stop When Claude finishes responding.
StopFailure When the turn ends due to an API error.
TeammateIdle When an agent team teammate is about to go idle.
InstructionsLoaded When a CLAUDE.md or .claude/rules/*.md file is loaded into context.
ConfigChange When a configuration file changes during a session.
CwdChanged When the working directory changes, for example when Claude executes a cd command.
DirectoryAdded When a working directory is added mid-session via /add-dir or the SDK register_repo_root control request.
FileChanged When a watched file changes on disk.
WorktreeCreate When a worktree is being created via --worktree, isolation: "worktree", or for a background session.
WorktreeRemove When a worktree that a WorktreeCreate hook created is being removed.
PreCompact Before context compaction.
PostCompact After context compaction completes.
PreModelSwitch Before Claude Code applies a model switch that you or a client requested.
PostModelSwitch After the session's model changes, including changes Claude Code makes on its own, such as restoring the model when you resume a session.
Elicitation When an MCP server requests user input during a tool call.
ElicitationResult After a user responds to an MCP elicitation, before the response is sent back to the server.
SessionEnd When a session terminates.
:::

Worked example: format every file Claude edits

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

Warning

Hooks run with your user permissions and no prompt. Only add hooks you've read and understand, and be careful with hooks that come inside repositories you cloned.

The hook receives details of the tool call as JSON on its input. jq pulls out the edited file's path, so Prettier formats only that file. You need jq and Prettier installed.

Environment variables

Set these in your shell or in the env block of settings.json. Most people never need more than a handful. Search this table when a guide or error message mentions one.

:::compact

Variable What it does
ANTHROPIC_API_KEY API key sent as X-Api-Key header.
ANTHROPIC_AUTH_TOKEN Custom value for the Authorization header (the value you set here will be prefixed with Bearer ).
ANTHROPIC_AWS_API_KEY Workspace API key for Claude Platform on AWS, generated in the AWS Console.
ANTHROPIC_AWS_BASE_URL Override the Claude Platform on AWS endpoint URL. Use for custom regions or when routing through an LLM…
ANTHROPIC_AWS_WORKSPACE_ID Required for Claude Platform on AWS. Sent on every request as the anthropic-workspace-id header.
ANTHROPIC_BASE_URL Override the API endpoint to route requests through a proxy or gateway.
ANTHROPIC_BEDROCK_BASE_URL Override the Amazon Bedrock endpoint URL. Use for custom Amazon Bedrock endpoints or when routing through…
ANTHROPIC_BEDROCK_MANTLE_BASE_URL Override the Amazon Bedrock Mantle endpoint URL. See Mantle endpoint.
ANTHROPIC_BEDROCK_REGION_PREFIX Cross-region inference profile prefix (us, eu, apac, jp, au, or global) Claude Code tries…
ANTHROPIC_BEDROCK_SERVICE_TIER Amazon Bedrock service tier (default, flex, or priority).
ANTHROPIC_BETAS Comma-separated list of additional anthropic-beta header values to include in API requests.
ANTHROPIC_CUSTOM_HEADERS Custom headers to add to requests (Name: Value format, newline-separated for multiple headers).
ANTHROPIC_CUSTOM_MODEL_OPTION Model ID to add as a custom entry in the /model picker.
ANTHROPIC_CUSTOM_MODEL_OPTION_DESCRIPTION Display description for the custom model entry in the /model picker.
ANTHROPIC_CUSTOM_MODEL_OPTION_NAME Display name for the custom model entry in the /model picker.
ANTHROPIC_CUSTOM_MODEL_OPTION_SUPPORTED_CAPABILITIES Comma-separated list of capabilities the custom model supports, for example effort,thinking.
ANTHROPIC_DEFAULT_FABLE_MODEL Model ID that the fable alias resolves to, and the ID Claude Code recognizes as a Fable model for…
ANTHROPIC_DEFAULT_FABLE_MODEL_DESCRIPTION Display description for the pinned Fable model in the /model picker.
ANTHROPIC_DEFAULT_FABLE_MODEL_NAME Display name for the pinned Fable model in the /model picker.
ANTHROPIC_DEFAULT_FABLE_MODEL_SUPPORTED_CAPABILITIES Comma-separated list of capabilities the pinned Fable model supports, for example effort,thinking.
ANTHROPIC_DEFAULT_HAIKU_MODEL Model ID that the haiku alias resolves to, also used for background functionality.
ANTHROPIC_DEFAULT_HAIKU_MODEL_DESCRIPTION Display description for the pinned Haiku model in the /model picker.
ANTHROPIC_DEFAULT_HAIKU_MODEL_NAME Display name for the pinned Haiku model in the /model picker.
ANTHROPIC_DEFAULT_HAIKU_MODEL_SUPPORTED_CAPABILITIES Comma-separated list of capabilities the pinned Haiku model supports, for example effort,thinking.
ANTHROPIC_DEFAULT_MODEL Model that new sessions start on by default.
ANTHROPIC_DEFAULT_OPUS_MODEL Model ID that the opus alias resolves to, and that opusplan uses while Plan Mode is active.
ANTHROPIC_DEFAULT_OPUS_MODEL_DESCRIPTION Display description for the pinned Opus model in the /model picker.
ANTHROPIC_DEFAULT_OPUS_MODEL_NAME Display name for the pinned Opus model in the /model picker.
ANTHROPIC_DEFAULT_OPUS_MODEL_SUPPORTED_CAPABILITIES Comma-separated list of capabilities the pinned Opus model supports, for example effort,thinking.
ANTHROPIC_DEFAULT_SONNET_MODEL Model ID that the sonnet alias resolves to, and that opusplan uses when Plan Mode is not active.
ANTHROPIC_DEFAULT_SONNET_MODEL_DESCRIPTION Display description for the pinned Sonnet model in the /model picker.
ANTHROPIC_DEFAULT_SONNET_MODEL_NAME Display name for the pinned Sonnet model in the /model picker.
ANTHROPIC_DEFAULT_SONNET_MODEL_SUPPORTED_CAPABILITIES Comma-separated list of capabilities the pinned Sonnet model supports, for example effort,thinking.
ANTHROPIC_FEDERATION_RULE_ID Federation rule ID for Workload Identity Federation.
ANTHROPIC_FOUNDRY_API_KEY API key for Microsoft Foundry authentication (see Microsoft Foundry).
ANTHROPIC_FOUNDRY_AUTH_TOKEN Bearer token for Microsoft Foundry authentication, such as a Microsoft Entra access token.
ANTHROPIC_FOUNDRY_BASE_URL Full base URL for the Microsoft Foundry resource (for example,…
ANTHROPIC_FOUNDRY_RESOURCE Microsoft Foundry resource name (for example, my-resource).
ANTHROPIC_MODEL Name of the model setting to use (see Model Configuration).
ANTHROPIC_ORGANIZATION_ID Organization ID for Workload Identity Federation.
ANTHROPIC_PROFILE Name of the Anthropic profile to authenticate with, such as one created by ant auth login or by signing…
ANTHROPIC_SMALL_FAST_MODEL [DEPRECATED] Name of Haiku-class model for background tasks.
ANTHROPIC_SMALL_FAST_MODEL_AWS_REGION Override AWS region for the Haiku-class model when using Amazon Bedrock or Amazon Bedrock Mantle.
ANTHROPIC_VERTEX_BASE_URL Override Google Cloud's Agent Platform endpoint URL. Use for custom Google Cloud's Agent Platform…
ANTHROPIC_VERTEX_PROJECT_ID GCP project ID that Google Cloud's Agent Platform requests are addressed to.
ANTHROPIC_WORKSPACE_ID Workspace ID for workload identity federation.
API_FORCE_IDLE_TIMEOUT Override the 5-minute body idle timeout that aborts a streaming model response when no bytes arrive.
API_TIMEOUT_MS Timeout for API requests in milliseconds (default: 600000, or 10 minutes; maximum: 2147483647).
AWS_BEARER_TOKEN_BEDROCK Amazon Bedrock API key for authentication (see Amazon Bedrock API keys).
BASH_DEFAULT_TIMEOUT_MS Default timeout for a foreground Bash or PowerShell tool command, in milliseconds (default: 120000, or 2…
BASH_MAX_OUTPUT_LENGTH Maximum number of characters of bash output that Claude Code reads back into a command's result (default:…
BASH_MAX_TIMEOUT_MS Maximum timeout the model can set for a foreground Bash or PowerShell tool command, in milliseconds…
BETA_TRACING_ENDPOINT OTLP/HTTP endpoint for detailed beta tracing: with ENABLE_BETA_TRACING_DETAILED=1, logs and traces go…
CCR_FORCE_BUNDLE Set to 1 to force claude --cloud to bundle and upload your local repository instead of cloning from…
CLAUDECODE Set to 1 in subprocesses Claude Code spawns (Bash and PowerShell tools, tmux sessions, hook commands,…
CLAUDE_AFK_COUNTDOWN_MS How many milliseconds before auto-continue the on-screen countdown appears on an unanswered…
CLAUDE_AFK_TIMEOUT_MS How many milliseconds of idle time before an unanswered AskUserQuestion dialog auto-continues without you.
CLAUDE_AGENT_SDK_DISABLE_BUILTIN_AGENTS Set to 1 to disable all built-in subagent types such as Explore and Plan.
CLAUDE_AGENT_SDK_MCP_NO_PREFIX Set to 1 to skip the mcp__<server>__ prefix on tool names from SDK-created MCP servers.
CLAUDE_ASYNC_AGENT_STALL_TIMEOUT_MS Stall timeout in milliseconds for subagents.
CLAUDE_AUTOCOMPACT_PCT_OVERRIDE Set the percentage (1-100) of the auto-compact window at which auto-compaction triggers.
CLAUDE_AUTO_BACKGROUND_TASKS Set to 1 to force-enable automatic backgrounding of long-running agent tasks.
CLAUDE_AX_PREPARK_MS In screen reader mode, how many milliseconds Claude Code waits before it writes a new or changed line.
CLAUDE_AX_SCREEN_READER Set to 1 to render screen-reader friendly output: flat text without decorative borders or animations.
CLAUDE_AX_STARTUP_QUIET_MS In screen reader mode, how many milliseconds Claude Code holds the first interface render after the…
CLAUDE_BASH_MAINTAIN_PROJECT_WORKING_DIR Return to the original working directory after each Bash or PowerShell command in the main session.
CLAUDE_BYTE_STREAM_IDLE_TIMEOUT_MS Timeout in milliseconds for the byte-level streaming idle watchdog; when set, it takes precedence over…
CLAUDE_CLIENT_PRESENCE_FILE Path to a file that an external tool, such as a screen-lock listener, creates when you unlock your screen…
CLAUDE_CODE_ACCESSIBILITY Set to 1 to keep the native terminal cursor visible and disable the inverted-text cursor indicator.
CLAUDE_CODE_ADDITIONAL_DIRECTORIES_CLAUDE_MD Set to 1 to load memory files from directories specified with --add-dir.
CLAUDE_CODE_ALT_SCREEN_FULL_REPAINT Set to 1 to repaint the entire screen on every frame in fullscreen rendering instead of sending…
CLAUDE_CODE_ALWAYS_ENABLE_EFFORT Set to 1 to send the effort parameter with every request, even when Claude Code does not recognize the…
CLAUDE_CODE_API_KEY_HELPER_TTL_MS Interval in milliseconds at which credentials should be refreshed (when using apiKeyHelper).
CLAUDE_CODE_ARTIFACT_AUTO_OPEN Set to 0 to stop Claude Code from opening the browser automatically when a new artifact is published.
CLAUDE_CODE_ARTIFACT_COMMENTS Set to 0 to stop Claude reading and replying to comments on an artifact.
CLAUDE_CODE_ARTIFACT_COMMENTS_AUTOREACT Set to 0 to stop Claude replying on its own to comments sent to it.
CLAUDE_CODE_ARTIFACT_PUBLISH_EDITS Set to 0 to stop Claude from applying small changes to an artifact's file as part of publishing it.
CLAUDE_CODE_ATTRIBUTION_HEADER Set to 0 to omit the attribution block, which carries the client version and a prompt fingerprint, from…
CLAUDE_CODE_AUTO_BACKGROUND_WORKER_CHECKIN_SECONDS Removed in v2.1.283. Use CLAUDE_CODE_WORKER_CHECKIN_SCHEDULE instead.
CLAUDE_CODE_AUTO_COMPACT_WINDOW Set the auto-compact window in tokens, from 100000 to 1000000.
CLAUDE_CODE_AUTO_CONNECT_IDE Override automatic IDE connection.
CLAUDE_CODE_AUTO_MODE_SERVER Controls whether Claude Code asks the server to review auto mode actions.
CLAUDE_CODE_AWS_CHAIN_RESOLVE_TIMEOUT_MS Time in milliseconds Claude Code waits for the AWS default credential provider chain to produce…
CLAUDE_CODE_BASH_EDIT_DIFF Set to 0 to turn off the diff of the files that changed while a Bash command ran, or 1 to record it in…
CLAUDE_CODE_BG_TASKS_REPORT_RUNNING Set to 0 to make a non-interactive session report an idle status to its host at every turn end, even…
CLAUDE_CODE_BRIDGE_SESSION_ID Set automatically in Bash tool and hook command subprocesses while the session has an active Remote…
CLAUDE_CODE_BS_AS_CTRL_BACKSPACE Set to 0 to make Claude Code read the 0x08 byte, also written ^H, as plain Backspace, or 1 to read…
CLAUDE_CODE_CERT_STORE Comma-separated list of CA certificate sources for TLS connections. bundled is the Mozilla CA set…
CLAUDE_CODE_CHILD_SESSION Set to 1 in subprocesses Claude Code spawns via the Bash, PowerShell, and Monitor tools, hook commands,…
CLAUDE_CODE_CLIENT_CERT Path to client certificate file for mTLS authentication.
CLAUDE_CODE_CLIENT_KEY Path to client private key file for mTLS authentication.
CLAUDE_CODE_CLIENT_KEY_PASSPHRASE Passphrase for encrypted CLAUDE_CODE_CLIENT_KEY (optional).
CLAUDE_CODE_CONNECT_TIMEOUT_MS Removed in v2.1.186 and now a no-op.
CLAUDE_CODE_DEBUG_LOGS_DIR Override the debug log file path.
CLAUDE_CODE_DEBUG_LOG_LEVEL Minimum log level written to the debug log file.
CLAUDE_CODE_DISABLE_1M_CONTEXT Set to 1 to turn off 1M context window support.
CLAUDE_CODE_DISABLE_ADAPTIVE_THINKING Set to 1 to disable adaptive reasoning on Opus 4.6 and Sonnet 4.6 and fall back to the fixed thinking…
CLAUDE_CODE_DISABLE_ADMIN_ENV_UNION Set to 1 to stop Claude Code from merging managed settings env blocks per key across admin sources, so…
CLAUDE_CODE_DISABLE_ADVISOR_TOOL Set to 1 to disable the advisor tool.
CLAUDE_CODE_DISABLE_AGENT_VIEW Set to 1 to turn off background agents and agent view: claude agents, --bg, /background, and the…
CLAUDE_CODE_DISABLE_ALTERNATE_SCREEN Set to 1 to disable fullscreen rendering and use the classic main-screen renderer.
CLAUDE_CODE_DISABLE_ARTIFACT Set to 1 to turn off the Artifact tool, which publishes session output as a private web page on claude.ai.
CLAUDE_CODE_DISABLE_ATTACHMENTS Set to 1 to disable attachment processing.
CLAUDE_CODE_DISABLE_AUTH_REFRESH_LOCK Set to 1 to make a Claude Code process run its gcpAuthRefresh or awsAuthRefresh command itself…
CLAUDE_CODE_DISABLE_AUTO_MEMORY Set to 1 to disable auto memory.
CLAUDE_CODE_DISABLE_BACKGROUND_TASKS Set to 1 to disable all background task functionality, including the run_in_background parameter on…
CLAUDE_CODE_DISABLE_BEDROCK_CONTENT_TYPE_DEFAULT Set to 1 to stop Claude Code from treating an Amazon Bedrock streaming response with a missing or empty…
CLAUDE_CODE_DISABLE_BEDROCK_CONTENT_TYPE_GUARD Set to 1 to skip the check that an Amazon Bedrock streaming response carries the…
CLAUDE_CODE_DISABLE_BG_EXIT_HANDOFF Set to 1 to stop a background session's running background shell commands, dynamic workflows, and, as of…
CLAUDE_CODE_DISABLE_BG_SHELL_PRESSURE_REAP Set to 1 to stop Claude Code from terminating background shell commands under memory pressure.
CLAUDE_CODE_DISABLE_BUNDLED_SKILLS Set to 1 to disable the skills and workflows included with Claude Code: bundled skills and workflows are…
CLAUDE_CODE_DISABLE_CFC_PROMPT Set to 1 to keep the Claude in Chrome browser tools available while omitting the Chrome section of the…
CLAUDE_CODE_DISABLE_CLAUDE_MDS Set to 1 to prevent loading any CLAUDE.md memory files into context, including user, project, and auto…
CLAUDE_CODE_DISABLE_CRON Set to 1 to disable scheduled tasks.
CLAUDE_CODE_DISABLE_DANGEROUS_RM_TIMEOUT Set to 1 to turn off the time limit on critical-path removal prompts.
CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS Set to 1 to strip pre-release anthropic-beta request headers, the body fields that pair with them, and…
CLAUDE_CODE_DISABLE_EXPLORE_PLAN_AGENTS Set to 1 to disable the built-in Explore and Plan subagents.
CLAUDE_CODE_DISABLE_FAST_MODE Set to 1 to disable fast mode.
CLAUDE_CODE_DISABLE_FEEDBACK_SURVEY Set to 1 to disable the "How is Claude doing?" session quality surveys.
CLAUDE_CODE_DISABLE_FILE_CHECKPOINTING Set to 1 to disable file checkpointing.
CLAUDE_CODE_DISABLE_FILE_MENTIONS Set to 1 to stop Claude Code from expanding @ mentions of files and directories.
CLAUDE_CODE_DISABLE_GIT_INSTRUCTIONS Set to 1 to remove built-in commit and PR workflow instructions and the git status snapshot from…
CLAUDE_CODE_DISABLE_INLINE_SHELL_RM_PROMPT Set to 1 to stop Claude Code from reading scripts passed to a shell with -c, such as `bash -c 'rm -rf…
CLAUDE_CODE_DISABLE_LEGACY_MODEL_REMAP Set to 1 to prevent automatic remapping of Opus 4 and Opus 4.1 to the current Opus version on the…
CLAUDE_CODE_DISABLE_MODEL_ACCESS_FALLBACK Set to 1 to stop Claude Code on Amazon Bedrock and Google Cloud's Agent Platform from switching to an…
CLAUDE_CODE_DISABLE_MOUSE Set to 1 to disable mouse tracking in fullscreen rendering.
CLAUDE_CODE_DISABLE_MOUSE_CLICKS Set to 1 to disable click, drag, and hover handling in fullscreen rendering while keeping mouse-wheel…
CLAUDE_CODE_DISABLE_MTLS_RELOAD_ON_STALE_CONNECTION Set to 1 to stop Claude Code from re-reading the mTLS client certificate and key when an API request…
CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC Set to any non-empty value, such as 1, to disable nonessential network traffic: auto-updates, telemetry,…
CLAUDE_CODE_DISABLE_NONSTREAMING_FALLBACK Set to 1 to disable the non-streaming fallback when a streaming request fails mid-stream.
CLAUDE_CODE_DISABLE_NOTIFICATION_PRESENCE_CHECK Set to 1 to send the PushNotification tool's desktop notification even while you are typing in or…
CLAUDE_CODE_DISABLE_OFFICIAL_MARKETPLACE_AUTOINSTALL Set to 1 to disable automatic registration of the official plugin marketplace.
CLAUDE_CODE_DISABLE_PERMISSION_PROMPT_NOTIFY_HOOKS Set to 1 to stop Claude Code from running your Notification hooks for unanswered permission requests…
CLAUDE_CODE_DISABLE_POLICY_SKILLS Set to 1 to skip loading skills from the system-wide managed skills directory.
CLAUDE_CODE_DISABLE_POWERSHELL_CMD_RM_DENY Set to 1 to turn off the PowerShell tool check that denies the cmd built-ins rd, rmdir, del, and…
CLAUDE_CODE_DISABLE_REFUSAL_FALLBACK Set to 1 to turn off the automatic model switch when a safety classifier flags a request, the behavior…
CLAUDE_CODE_DISABLE_STRUCTURED_OUTPUTS Set to 1 to stop Claude Code from sending the structured-output output_config.format field and the…
CLAUDE_CODE_DISABLE_SUBSTITUTION_RM_PROMPT Set to 1 to turn off the critical-path check for a recursive rm whose target is entirely the output of…
CLAUDE_CODE_DISABLE_TERMINAL_TITLE Set to 1 to disable automatic terminal title updates based on conversation context.
CLAUDE_CODE_DISABLE_THINKING Set to 1 to omit the thinking parameter from API requests entirely.
CLAUDE_CODE_DISABLE_UNKNOWN_MODEL_WINDOW_ENFORCEMENT Set to 1 to skip proactive auto-compaction when Claude Code doesn't recognize the model ID, such as an…
CLAUDE_CODE_DISABLE_VIRTUAL_SCROLL Set to 1 to disable virtual scrolling in fullscreen rendering and render every message in the transcript.
CLAUDE_CODE_DISABLE_WEB_FETCH Set to 1 to turn off the WebFetch tool.
CLAUDE_CODE_DISABLE_WINDOWS_SHELL_LAUNCHER Set to 1 to start PowerShell tool commands on Windows directly instead of through the cmd.exe launcher.
CLAUDE_CODE_DISABLE_WORKFLOWS Set to 1 to disable workflows.
CLAUDE_CODE_EFFORT_LEVEL Set the effort level for supported models.
CLAUDE_CODE_EMIT_SESSION_STATE_EVENTS Set to 1 to add session_state_changed messages, which carry the session's state, to the message stream.
CLAUDE_CODE_ENABLE_AUTO_MODE Accepted for compatibility with older releases and has no effect.
CLAUDE_CODE_ENABLE_AWAY_SUMMARY Override session recap availability.
CLAUDE_CODE_ENABLE_BACKGROUND_PLUGIN_REFRESH Set to 1 to refresh plugin state at turn boundaries in non-interactive mode after a background install…
CLAUDE_CODE_ENABLE_CFC Set to 1 to start a CLI session with Chrome integration on, or 0 to start with it off.
CLAUDE_CODE_ENABLE_FEEDBACK_SURVEY_FOR_OTEL Set to 1 to route the "How is Claude doing?" session quality survey to your own OpenTelemetry collector…
CLAUDE_CODE_ENABLE_FINE_GRAINED_TOOL_STREAMING Controls whether tool call inputs stream from the API as Claude generates them.
CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY Set to 1 to populate the /model picker from your gateway's /v1/models endpoint when…
CLAUDE_CODE_ENABLE_OPUS_4_7_FAST_MODE Removed in v2.1.142, when the fast mode default moved from Opus 4.6 to Opus 4.7.
CLAUDE_CODE_ENABLE_PROMPT_SUGGESTION Set to false to turn off prompt suggestions, the grayed-out predictions that appear in your prompt input.
CLAUDE_CODE_ENABLE_TASKS Selects which task-tracking tools Claude Code provides in sessions that have them.
CLAUDE_CODE_ENABLE_TELEMETRY Set to 1 to enable OpenTelemetry data collection for metrics and logging.
CLAUDE_CODE_ENABLE_TODO_TOOLS Set to 1 to get the task-tracking tools on every model.
CLAUDE_CODE_EXIT_AFTER_STOP_DELAY Time in milliseconds to wait after the query loop becomes idle before automatically exiting.
CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS Set to 1 to enable agent teams.
CLAUDE_CODE_EXTRA_BODY JSON object to merge into the top level of every API request body.
CLAUDE_CODE_FILE_READ_MAX_OUTPUT_TOKENS Override the default token limit for file reads, which is 25,000 tokens.
CLAUDE_CODE_FORCE_SESSION_PERSISTENCE Set to 1 to force transcript persistence, prompt history, and claude agents registration even when…
CLAUDE_CODE_FORCE_STRIKETHROUGH Set to 1 to force strikethrough rendering for ~~text~~ in Claude's responses when your terminal…
CLAUDE_CODE_FORCE_SYNC_OUTPUT Set to 1 to force-enable DEC private mode 2026 synchronized output when your terminal supports it but is…
CLAUDE_CODE_FORCE_TERMINAL_IMAGES Set to 1 to draw mod Image elements as pictures when your terminal draws kitty graphics protocol…
CLAUDE_CODE_FORK_SUBAGENT Controls fork mode, which lets Claude spawn forked subagents itself and is on by default in interactive…
CLAUDE_CODE_FORWARD_SUBAGENT_TEXT Set to 1 to emit subagent text and thinking blocks in claude -p --output-format stream-json output,…
CLAUDE_CODE_GATEWAY_HINT_HEADERS Set to 1 to send the gateway hint headers, such as x-claude-code-request-class and…
CLAUDE_CODE_GATEWAY_MODEL_DISCOVERY_TIMEOUT_MS Timeout in milliseconds for the gateway model discovery request that…
CLAUDE_CODE_GIT_BASH_PATH Windows only: path to the Git Bash executable (bash.exe).
CLAUDE_CODE_GLOB_HIDDEN Set to false to exclude dotfiles from results when Claude invokes the Glob tool.
CLAUDE_CODE_GLOB_NO_IGNORE Set to false to make the Glob tool respect .gitignore patterns.
CLAUDE_CODE_GLOB_TIMEOUT_SECONDS Timeout in seconds for Glob tool file discovery.
CLAUDE_CODE_GOAL_CHECKIN_MINUTES How many minutes background work can keep an active goal waiting before Claude Code asks Claude to check…
CLAUDE_CODE_GZIP_REQUEST_BODIES Set to 0 to turn off gzip compression of Claude API, telemetry, and artifact publish request bodies sent…
CLAUDE_CODE_HIDE_CWD Set to 1 to hide the working directory in the startup logo.
CLAUDE_CODE_IDE_HOST_OVERRIDE Override the host address used to connect to the IDE extension.
CLAUDE_CODE_IDE_SKIP_AUTO_INSTALL Set to 1 to skip auto-installation of IDE extensions.
CLAUDE_CODE_IDE_SKIP_VALID_CHECK Set to 1 to skip validation of IDE lockfile entries during connection.
CLAUDE_CODE_IDLE_COMPACT_MIN_TOKENS Context size, in tokens, at which idle compaction applies to a conversation (default: 200000).
CLAUDE_CODE_MAX_CONCURRENT_SUBAGENTS How many subagents can be running in one session before the Agent tool refuses to spawn another (default: 20).
CLAUDE_CODE_MAX_CONTEXT_TOKENS Override the context window size Claude Code assumes for the active model.
CLAUDE_CODE_MAX_MCP_DESCRIPTION_LENGTH Maximum length in characters of MCP tool descriptions and MCP server instructions that Claude Code sends…
CLAUDE_CODE_MAX_OUTPUT_TOKENS Set the maximum number of output tokens for most requests.
CLAUDE_CODE_MAX_RETRIES Override the number of times to retry failed API requests (default: 10).
CLAUDE_CODE_MAX_SUBAGENTS_PER_SESSION Removed in v2.1.224 and now a no-op.
CLAUDE_CODE_MAX_SUBAGENT_SPAWN_DEPTH Number of subagent layers allowed below the main conversation (default: 3).
CLAUDE_CODE_MAX_TOOL_USE_CONCURRENCY Maximum number of read-only tools and subagents that can execute in parallel (default: 10).
CLAUDE_CODE_MAX_TURNS Cap the number of agentic turns when no explicit limit is passed.
CLAUDE_CODE_MAX_WEB_SEARCHES_PER_SESSION Cap on WebSearch calls (default: 200).
CLAUDE_CODE_MCP_ALLOWLIST_ENV Set to 1 to spawn stdio MCP servers with only a safe baseline environment plus the server's configured…
CLAUDE_CODE_MCP_AUTO_BACKGROUND_MS Elapsed time in milliseconds before a still-running MCP tool call moves to a background task (default:…
CLAUDE_CODE_MCP_STARTUP_WAIT_MS How long in milliseconds the first turn of a non-interactive session waits for MCP servers that are still…
CLAUDE_CODE_MCP_TOOL_IDLE_TIMEOUT Idle timeout in milliseconds for MCP tool calls.
CLAUDE_CODE_MESSAGING_SOCKET Set by Claude Code, not by you: in sessions that bind an inbox socket, Claude Code exports that socket's…
CLAUDE_CODE_MESSAGING_TOKEN Set by Claude Code, not by you: in sessions that bind an inbox socket, Claude Code exports this…
CLAUDE_CODE_NATIVE_CURSOR Set to 1 to show the terminal's own cursor at the input caret instead of a drawn block.
CLAUDE_CODE_NEW_INIT Set to 1 to make /init run an interactive setup flow.
CLAUDE_CODE_NONBLOCKING_STDOUT Set to 1 to write terminal output through a second non-blocking file descriptor, so a terminal that…
CLAUDE_CODE_NONSTREAMING_TIMEOUT_RETRIES Limit how many times Claude Code re-sends a non-streaming request that times out.
CLAUDE_CODE_NO_FLICKER Set to 1 to enable fullscreen rendering, a research preview that reduces flicker and keeps memory flat…
CLAUDE_CODE_OAUTH_REFRESH_TOKEN OAuth refresh token for Claude.ai authentication.
CLAUDE_CODE_OAUTH_SCOPES Space-separated OAuth scopes the refresh token was issued with, such as `"user:profile user:inference…
CLAUDE_CODE_OAUTH_TOKEN OAuth access token for claude.ai authentication.
CLAUDE_CODE_OPUS_4_6_FAST_MODE_OVERRIDE Removed in v2.1.160 and now a no-op.
CLAUDE_CODE_OTEL_CONTENT_MAX_LENGTH Maximum length of content-bearing OpenTelemetry attributes (model responses, tool content, system prompts,…
CLAUDE_CODE_OTEL_DIAG_STDERR Set to 1 to write OpenTelemetry exporter diagnostic errors to stderr.
CLAUDE_CODE_OTEL_FLUSH_TIMEOUT_MS Timeout in milliseconds for flushing pending OpenTelemetry spans (default: 5000).
CLAUDE_CODE_OTEL_HEADERS_HELPER_DEBOUNCE_MS Interval for refreshing dynamic OpenTelemetry headers in milliseconds (default: 1740000 / 29 minutes).
CLAUDE_CODE_OTEL_SHUTDOWN_TIMEOUT_MS Timeout in milliseconds for the OpenTelemetry exporter to finish on shutdown (default: 2000).
CLAUDE_CODE_OVERLOADED_RETRY_BASE_DELAY_MS Starting delay in milliseconds, in place of the default 500, of the exponential backoff between automatic…
CLAUDE_CODE_OVERLOADED_RETRY_MAX_DELAY_MS Ceiling in milliseconds, in place of the default 32000, on the exponential backoff between automatic…
CLAUDE_CODE_PACKAGE_MANAGER_AUTO_UPDATE Set to 1 to let Claude Code run your package manager's upgrade command in the background when a new…
CLAUDE_CODE_PERFORCE_MODE Set to 1 to enable Perforce-aware write protection.
CLAUDE_CODE_PLUGIN_CACHE_DIR Override the plugins root directory.
CLAUDE_CODE_PLUGIN_DIRS Plugin directories to load for the session, each loaded the way a --plugin-dir flag loads it.
CLAUDE_CODE_PLUGIN_DIR_WATCH Controls whether Claude Code reloads a mod when the mod's files change.
CLAUDE_CODE_PLUGIN_GIT_TIMEOUT_MS Timeout in milliseconds for cloning or refreshing a plugin marketplace (default: 120000).
CLAUDE_CODE_PLUGIN_KEEP_MARKETPLACE_ON_FAILURE Set to 1 to skip the re-clone attempt and keep using the existing marketplace checkout when a…
CLAUDE_CODE_PLUGIN_PREFER_HTTPS Set to 1 to clone GitHub owner/repo shorthand sources over HTTPS instead of SSH. Applies to plugin…
CLAUDE_CODE_PLUGIN_SEED_DIR Path to one or more read-only plugin seed directories, separated by : on Unix or ; on Windows.
CLAUDE_CODE_POWERSHELL_RESPECT_EXECUTION_POLICY Set to 1 to stop Claude Code from passing -ExecutionPolicy Bypass when spawning PowerShell for tool…
CLAUDE_CODE_PRINT_BG_WAIT_CEILING_MS Ceiling in milliseconds on idle waiting for background work, such as subagents and workflows, after the…
CLAUDE_CODE_PROCESS_WRAPPER Launch the processes Claude Code starts from its own binary, such as the background service that hosts…
CLAUDE_CODE_PROJECT_DIR_NAME Set together with CLAUDE_CONFIG_DIR to choose the projects/ directory name Claude Code stores that…
CLAUDE_CODE_PROMPT_CACHE_TTL Set 5m or 1h, the only values Claude Code accepts, to choose the prompt cache TTL for the main…
CLAUDE_CODE_PROPAGATE_TRACEPARENT Set to 1 to propagate W3C trace context when ANTHROPIC_BASE_URL points at a custom proxy.
CLAUDE_CODE_PROVIDER_MANAGED_BY_HOST Set by host platforms that embed Claude Code and manage model provider routing on its behalf.
CLAUDE_CODE_PROXY_RESOLVES_HOSTS Set to 1 to allow the proxy to perform DNS resolution instead of the caller.
CLAUDE_CODE_REMOTE Set automatically to true when Claude Code is running as a cloud session.
CLAUDE_CODE_REMOTE_SESSION_ID Set automatically in cloud sessions to the current session's ID. Read this to construct a link back to the…
CLAUDE_CODE_RESTRICTED Set to 1 to start the session in restricted mode, the same as passing --restricted.
CLAUDE_CODE_RESUME_INTERRUPTED_TURN Set to 1 to automatically resume if the previous session ended mid-turn.
CLAUDE_CODE_RESUME_INTERRUPTED_TURN_MAX_AGE_MS Maximum age in milliseconds of the last transcript message for a session that ended mid-turn to continue…
CLAUDE_CODE_RESUME_PROMPT Override the continuation message Claude Code sends to Claude when CLAUDE_CODE_RESUME_INTERRUPTED_TURN…
CLAUDE_CODE_RETRY_WATCHDOG Set to 1 for unattended sessions such as eval harnesses, CI jobs, or remote workers.
CLAUDE_CODE_RETRY_WATCHDOG_MAX_WAIT_MS Maximum time in milliseconds that each API request spends waiting out 429 and 529 errors when…
CLAUDE_CODE_SAFE_MODE Set to 1 to start in safe mode: CLAUDE.md, skills, plugins, hooks, MCP servers, custom commands and…
CLAUDE_CODE_SCRIPT_CAPS JSON object limiting how many times specific scripts may be invoked per session when…
CLAUDE_CODE_SCROLL_SPEED Set the mouse wheel scroll multiplier in fullscreen rendering.
CLAUDE_CODE_SEND_FEEDBACK Set to 0 to turn off Claude-drafted feedback for a session.
CLAUDE_CODE_SESSIONEND_HOOKS_TIMEOUT_MS Override the time budget in milliseconds for SessionEnd hooks.
CLAUDE_CODE_SESSION_ID Set automatically to the current session ID in Bash and PowerShell tool subprocesses, hook command…
CLAUDE_CODE_SHELL Set the shell Claude Code uses to run Bash tool commands.
CLAUDE_CODE_SHELL_PREFIX Command prefix that wraps shell commands Claude Code spawns: Bash tool calls, hook commands, status line…
CLAUDE_CODE_SIMPLE Set to 1 to run with a minimal system prompt and only the Bash, file read, and file edit tools.
CLAUDE_CODE_SIMPLE_SYSTEM_PROMPT Choose between Claude Code's full system prompt and a shorter one with abbreviated tool descriptions.
CLAUDE_CODE_SKIP_ANTHROPIC_AWS_AUTH Skip client-side authentication for Claude Platform on AWS, for gateways that sign requests themselves.
CLAUDE_CODE_SKIP_AWS_CRED_CACHE Set to 1 to turn off the in-process cache of credentials resolved from the AWS default credential…
CLAUDE_CODE_SKIP_BEDROCK_AUTH Skip AWS authentication for Amazon Bedrock (for example, when using an LLM gateway).
CLAUDE_CODE_SKIP_FAST_MODE_NETWORK_ERRORS Set to 1 to treat a failed fast mode availability check as available, for networks that block the…
CLAUDE_CODE_SKIP_FAST_MODE_ORG_CHECK Set to 1 to skip the client-side fast mode availability check, for proxies that intercept the check's…
CLAUDE_CODE_SKIP_FOUNDRY_AUTH Skip Azure authentication for Microsoft Foundry, for a proxy or gateway that injects its own…
CLAUDE_CODE_SKIP_MANTLE_AUTH Skip AWS authentication for Amazon Bedrock Mantle (for example, when using an LLM gateway).
CLAUDE_CODE_SKIP_MODEL_ACCESS_MEMORY The startup model checks on Amazon Bedrock and Google Cloud's Agent Platform remember on this machine…
CLAUDE_CODE_SKIP_PROMPT_HISTORY Set to 1 to skip writing prompt history and session transcripts to disk.
CLAUDE_CODE_SKIP_VERTEX_AUTH Skip Google authentication for Google Cloud's Agent Platform (for example, when using an LLM gateway).
CLAUDE_CODE_STARTUP_FAILURE_RESULTS Set to 1 to have a session started with --output-format stream-json write a result message naming why…
CLAUDE_CODE_STOP_HOOK_BLOCK_CAP Maximum number of consecutive times a Stop or SubagentStop hook may block the turn from ending before…
CLAUDE_CODE_SUBAGENT_MODEL The default model for subagents, agent team teammates, and workflow agents that aren't assigned a model…
CLAUDE_CODE_SUBAGENT_MODEL_FORCE Set to 1 to force one model onto subagents, teammates, and workflow agents.
CLAUDE_CODE_SUBAGENT_PROMPT_CACHE_TTL Set 5m or 1h, the only values Claude Code accepts, to choose the prompt cache TTL for requests outside…
CLAUDE_CODE_SUBPROCESS_ENV_SCRUB Set to 1 to strip credentials from the environments of the subprocesses Claude Code starts, such as Bash…
CLAUDE_CODE_SYNC_PLUGIN_INSTALL Set to 1 in non-interactive mode (the -p flag) to wait for plugin installation to complete before the…
CLAUDE_CODE_SYNC_PLUGIN_INSTALL_TIMEOUT_MS Timeout in milliseconds for synchronous plugin installation.
CLAUDE_CODE_SYNC_SKILLS Set to 1 in non-interactive mode with the -p flag to make Claude Code download the skills enabled for…
CLAUDE_CODE_SYNC_SKILLS_INSTALL_TIMEOUT_MS Timeout in milliseconds for the skills resync that runs mid-session when an app built on the Agent SDK…
CLAUDE_CODE_SYNC_SKILLS_WAIT_TIMEOUT_MS Timeout in milliseconds for the first query to wait for the initial skill list when…
CLAUDE_CODE_SYNTAX_HIGHLIGHT Set to false to disable syntax highlighting in diff output.
CLAUDE_CODE_TASK_LIST_ID Share a task list across sessions.
CLAUDE_CODE_TEAM_TEARDOWN_PARK_TIMEOUT_MS Override, in milliseconds, how long a non-interactive session waits at exit for its agent team to finish…
CLAUDE_CODE_TMPDIR Override the temp directory used for internal temp files.
CLAUDE_CODE_TMUX_TRUECOLOR Set to any non-empty value, such as 1, to allow 24-bit truecolor output inside tmux. **Setting it to 0…
CLAUDE_CODE_TOOL_MEMORY_CGROUP_EXCLUDE On Linux and WSL, set to a comma-separated list of the kinds of processes Claude Code excludes from the…
CLAUDE_CODE_TOOL_MEMORY_LIMIT On Linux and WSL, set to a size such as 4G to cap the memory that Bash and PowerShell tool commands can…
CLAUDE_CODE_TRANSCRIPT_LOCAL_GC Set to 1 to limit how large the transcript file of a long -p or Agent SDK session grows.
CLAUDE_CODE_USER_DIALOG_TIMEOUT_MS Deadline in milliseconds before Claude Code cancels a dialog it forwards to a remote client such as a…
CLAUDE_CODE_USE_ANTHROPIC_AWS Use Claude Platform on AWS.
CLAUDE_CODE_USE_BEDROCK Use Amazon Bedrock.
CLAUDE_CODE_USE_FOUNDRY Use Microsoft Foundry.
CLAUDE_CODE_USE_MANTLE Use the Amazon Bedrock Mantle endpoint.
CLAUDE_CODE_USE_NATIVE_FILE_SEARCH Set to 1 to discover custom commands, subagents, and output styles using Node.js file APIs instead of…
CLAUDE_CODE_USE_POWERSHELL_TOOL Controls the PowerShell tool.
CLAUDE_CODE_USE_VERTEX Use Google Cloud's Agent Platform.
CLAUDE_CODE_WEBFETCH_CACHE_TTL_MS Set to the number of milliseconds WebFetch keeps each fetched URL's response cached.
CLAUDE_CODE_WEBFETCH_DEADLINE_MS Upper bound in milliseconds on how long WebFetch waits for a page to download, including any redirects it…
CLAUDE_CODE_WEB_SEARCH_REFILLS_PER_HOUR Rate at which a session's WebSearch limit refills, in calls per hour.
CLAUDE_CODE_WORKER_CHECKIN_SCHEDULE When CLAUDE_AUTO_BACKGROUND_TASKS is set to 1, how long Claude Code waits before each reminder to…
CLAUDE_CODE_WORKFLOW_MAX_CONCURRENT_AGENTS How many agents a single workflow run executes at once, from 1 to 256.
CLAUDE_CODE_WORKFLOW_PREFIX_STAGGER_MS Upper bound in milliseconds on how long a workflow agent waits for a same-prefix sibling's first response…
CLAUDE_CODE_YOU_SHOULD_KNOW_DEBUG Set to 1 to show, above the prompt, a line for each of the last three checks the You should know mod…
CLAUDE_CONFIG_DIR Override the configuration directory (default: ~/.claude).
CLAUDE_DISABLE_ADOPT Set to 1 to stop in-flight background work instead of carrying it over when you background a session by…
CLAUDE_EFFORT Set automatically in Bash tool subprocesses and hook commands to the effort level in effect when the…
CLAUDE_ENABLE_BYTE_WATCHDOG Set to 1 to force-enable the byte-level streaming idle watchdog, or set to 0 to force-disable it. 0…
CLAUDE_ENABLE_BYTE_WATCHDOG_BEDROCK Set to 1 to enable the byte-level streaming idle watchdog on Amazon Bedrock vnd.amazon.eventstream…
CLAUDE_ENABLE_STREAM_WATCHDOG Set to 0 to force-disable the event-level streaming idle watchdog, or set to 1 to force-enable it.
CLAUDE_ENV_FILE Path to a shell script whose contents Claude Code runs before each Bash command in the same shell process,…
CLAUDE_JOB_DIR Set by Claude Code in each background session to that session's ~/.claude/jobs/<id> directory.
CLAUDE_PID Claude Code sets this to its own process ID in the subprocesses it spawns: Bash and PowerShell tool…
CLAUDE_REMOTE_CONTROL_SESSION_NAME_PREFIX Prefix for auto-generated Remote Control session names when no explicit name is provided.
CLAUDE_STREAM_FIRST_BYTE_TIMEOUT_MS Deadline in milliseconds for the first response byte of a streaming request, on the connections where the…
CLAUDE_STREAM_IDLE_TIMEOUT_MS Timeout in milliseconds before the event- and byte-level streaming idle watchdogs close a stalled connection.
CLAUDE_SUBAGENT_BG_SHELL_MAX_MS Removed in v2.1.260 and now a no-op.
DEBUG Set to 1 to enable debug mode, equivalent to launching with --debug.
DISABLE_AUTOUPDATER Set to 1 to disable automatic background updates.
DISABLE_AUTO_COMPACT Set to 1 to disable automatic compaction when approaching the context limit, along with idle compaction.
DISABLE_COMPACT Set to 1 to disable all compaction: both automatic compaction and the manual /compact command.
DISABLE_COST_WARNINGS Set to 1 to disable cost warning messages.
DISABLE_DOCTOR_COMMAND Set to 1 to hide the /doctor setup checkup skill and its /checkup alias.
DISABLE_ERROR_REPORTING Set to any non-empty value, such as 1, to opt out of error reporting. **Setting it to 0 or false…
DISABLE_EXTRA_USAGE_COMMAND Set to 1 to hide the /usage-credits command that lets users purchase additional usage beyond rate limits.
DISABLE_FEEDBACK_COMMAND Set to 1 to disable the /feedback command and Claude-drafted feedback.
DISABLE_GROWTHBOOK Set to 1 or true to disable GrowthBook feature-flag fetching and use code defaults for every flag.
DISABLE_INSTALLATION_CHECKS Set to 1 to disable installation warnings.
DISABLE_INSTALL_GITHUB_APP_COMMAND Set to 1 to hide the /install-github-app command.
DISABLE_INTERLEAVED_THINKING Set to 1 to prevent sending the interleaved-thinking beta header.
DISABLE_LOGIN_COMMAND Set to 1 to hide the /login command.
DISABLE_LOGOUT_COMMAND Set to 1 to hide the /logout command.
DISABLE_PROMPT_CACHING Set to 1 to disable prompt caching for all models (takes precedence over per-model settings).
DISABLE_PROMPT_CACHING_FABLE Set to 1 to disable prompt caching for Fable models.
DISABLE_PROMPT_CACHING_HAIKU Set to 1 to disable prompt caching for the default Haiku model, wherever it runs.
DISABLE_PROMPT_CACHING_OPUS Set to 1 to disable prompt caching for the default Opus model.
DISABLE_PROMPT_CACHING_SONNET Set to 1 to disable prompt caching for the default Sonnet model.
DISABLE_TELEMETRY Set to any non-empty value, such as 1, to opt out of telemetry. **Setting it to 0 or false still…
DISABLE_UPDATES Set to 1 to block all updates including manual claude update and claude install.
DISABLE_UPGRADE_COMMAND Set to 1 to hide the /upgrade command.
DO_NOT_TRACK Set to 1 to opt out of telemetry, with the same effect as DISABLE_TELEMETRY, including on feature-flag…
ENABLE_BETA_TRACING_DETAILED Set to 1, and set BETA_TRACING_ENDPOINT to your OTLP/HTTP collector endpoint, to turn on detailed beta…
ENABLE_CLAUDEAI_MCP_SERVERS Set to false to stop Claude Code from fetching claude.ai MCP servers.
ENABLE_PROMPT_CACHING_1H Set to 1 to request a 1-hour prompt cache TTL instead of the default 5 minutes.
ENABLE_PROMPT_CACHING_1H_BEDROCK Deprecated.
ENABLE_TOOL_SEARCH Controls MCP tool search.
FALLBACK_FOR_ALL_PRIMARY_MODELS Set to any non-empty value, such as 1, to make Claude Code stop retrying on repeated overload errors for…
FORCE_AUTOUPDATE_PLUGINS Set to 1 to force plugin auto-updates even when the main auto-updater is disabled via DISABLE_AUTOUPDATER.
FORCE_HYPERLINK Set to 1 to enable clickable OSC 8 hyperlinks when your terminal supports them but isn't auto-detected,…
FORCE_PROMPT_CACHING_5M Set to 1 to force the 5-minute prompt cache TTL even when 1-hour TTL would otherwise apply.
HTTP_PROXY Specify HTTP proxy server for network connections.
HTTPS_PROXY Specify HTTPS proxy server for network connections.
IS_DEMO Set to any non-empty value, such as 1, to enable demo mode: hides your email and organization name from…
MAX_MCP_OUTPUT_TOKENS Maximum number of tokens allowed in MCP tool responses (default: 25000).
MAX_STRUCTURED_OUTPUT_RETRIES Number of attempts Claude Code allows when the model's response fails validation against the…
MAX_THINKING_TOKENS Fixed token budget for extended thinking.
MCP_CLIENT_SECRET OAuth client secret for MCP servers that require pre-configured credentials.
MCP_CONNECTION_NONBLOCKING Controls whether startup waits for MCP servers to connect before the first query.
MCP_CONNECT_TIMEOUT_MS How long blocking MCP startup waits, in milliseconds, for the connection batch before snapshotting the…
MCP_DISCOVERY_CACHE Turns the MCP discovery cache on or off.
MCP_DISCOVERY_CACHE_MAX_STALE_S Maximum age, in seconds, of a discovery-cache entry (default: 14400, or 4 hours).
MCP_DISCOVERY_CACHE_STRIKES At a start where a discovery-cache entry is older than MCP_DISCOVERY_CACHE_TTL_S, Claude Code refreshes…
MCP_DISCOVERY_CACHE_TTL_S Seconds for which Claude Code uses a discovery-cache entry without refreshing it (default: 900).
MCP_OAUTH_CALLBACK_PORT Fixed port for the OAuth redirect callback, as an alternative to --callback-port when adding an MCP…
MCP_PROTOCOL_NEGOTIATION On the v2 MCP client runtime only, whether Claude Code probes servers for MCP protocol revision…
MCP_REMOTE_SERVER_CONNECTION_BATCH_SIZE Maximum number of remote MCP servers (HTTP/SSE) to connect in parallel during startup (default: 20).
MCP_SDK_GENERATION Pin which MCP client runtime this process connects to MCP servers with: v1, built on MCP TypeScript SDK…
MCP_SERVER_CONNECTION_BATCH_SIZE Maximum number of local MCP servers (stdio) to connect in parallel during startup (default: 3).
MCP_TIMEOUT Timeout in milliseconds for MCP server startup (default: 30000, or 30 seconds).
MCP_TOOL_TIMEOUT Timeout in milliseconds for MCP tool execution (default: 100000000, about 28 hours).
NO_PROXY List of domains and IPs to which requests will be directly issued, bypassing proxy.
OTEL_ATTRIBUTE_VALUE_LENGTH_LIMIT Standard OpenTelemetry SDK limit on attribute value length.
OTEL_LOG_ASSISTANT_RESPONSES Set to 1 to include the model's response text on assistant_response OpenTelemetry log events.
OTEL_LOG_MANAGED_SETTINGS Set to 1 to add the redacted managed settings, and a SHA-256 digest of the settings before redaction, to…
OTEL_LOG_RAW_API_BODIES Emit Anthropic Messages API request and response JSON as api_request_body / api_response_body log events.
OTEL_LOG_TOOL_CONTENT Set to 1 to include tool content in the tool.output OpenTelemetry span event.
OTEL_LOG_TOOL_DETAILS Set to 1 to include tool input arguments; MCP server names; user-authored workflow names; raw error…
OTEL_LOG_USER_PROMPTS Set to 1 to include user prompt text in OpenTelemetry traces and logs.
OTEL_METRICS_INCLUDE_ACCOUNT_UUID Set to false to exclude account UUID from metrics attributes (default: included).
OTEL_METRICS_INCLUDE_ENTRYPOINT Set to true to include the session entrypoint in metrics attributes (default: excluded).
OTEL_METRICS_INCLUDE_REPOSITORY Set to true to tag OpenTelemetry metrics and events with vcs.* attributes identifying the session's…
OTEL_METRICS_INCLUDE_RESOURCE_ATTRIBUTES As of v2.1.161, Claude Code attaches OTEL_RESOURCE_ATTRIBUTES keys to metric datapoint labels.
OTEL_METRICS_INCLUDE_SESSION_ID Set to false to exclude session ID from metrics attributes (default: included).
OTEL_METRICS_INCLUDE_VERSION Set to true to include Claude Code version in metrics attributes (default: excluded).
SLASH_COMMAND_TOOL_CHAR_BUDGET Override the character budget for skill metadata shown to the Skill tool.
TASK_MAX_OUTPUT_LENGTH Removed in v2.1.277 and now a no-op, together with the TaskOutput tool it sized.
USE_BUILTIN_RIPGREP Set to 0 to use system-installed rg instead of rg included with Claude Code.
VERTEX_REGION_CLAUDE_3_5_HAIKU Override region for Claude 3.5 Haiku when using Google Cloud's Agent Platform.
VERTEX_REGION_CLAUDE_3_5_SONNET Override region for Claude 3.5 Sonnet when using Google Cloud's Agent Platform.
VERTEX_REGION_CLAUDE_3_7_SONNET Override region for Claude 3.7 Sonnet when using Google Cloud's Agent Platform.
VERTEX_REGION_CLAUDE_4_0_OPUS Override region for Claude Opus 4 when using Google Cloud's Agent Platform.
VERTEX_REGION_CLAUDE_4_0_SONNET Override region for Claude Sonnet 4 when using Google Cloud's Agent Platform.
VERTEX_REGION_CLAUDE_4_1_OPUS Override region for Claude Opus 4.1 when using Google Cloud's Agent Platform.
VERTEX_REGION_CLAUDE_4_5_OPUS Override region for Claude Opus 4.5 when using Google Cloud's Agent Platform.
VERTEX_REGION_CLAUDE_4_5_SONNET Override region for Claude Sonnet 4.5 when using Google Cloud's Agent Platform.
VERTEX_REGION_CLAUDE_4_6_OPUS Override region for Claude Opus 4.6 when using Google Cloud's Agent Platform.
VERTEX_REGION_CLAUDE_4_6_SONNET Override region for Claude Sonnet 4.6 when using Google Cloud's Agent Platform.
VERTEX_REGION_CLAUDE_4_7_OPUS Override region for Claude Opus 4.7 when using Google Cloud's Agent Platform.
VERTEX_REGION_CLAUDE_4_8_OPUS Override region for Claude Opus 4.8 when using Google Cloud's Agent Platform.
VERTEX_REGION_CLAUDE_5_5_OPUS Override region for Claude Opus 5.5 when using Google Cloud's Agent Platform.
VERTEX_REGION_CLAUDE_5_5_SONNET Override region for Claude Sonnet 5.5 when using Google Cloud's Agent Platform.
VERTEX_REGION_CLAUDE_5_OPUS Override region for Claude Opus 5 when using Google Cloud's Agent Platform.
VERTEX_REGION_CLAUDE_5_SONNET Override region for Claude Sonnet 5 when using Google Cloud's Agent Platform.
VERTEX_REGION_CLAUDE_FABLE_5 Override region for Claude Fable 5 when using Google Cloud's Agent Platform.
VERTEX_REGION_CLAUDE_FABLE_5_1 Override region for Claude Fable 5.1 when using Google Cloud's Agent Platform.
VERTEX_REGION_CLAUDE_HAIKU_4_5 Override region for Claude Haiku 4.5 when using Google Cloud's Agent Platform.
VERTEX_REGION_CLAUDE_HAIKU_5_5 Override region for Claude Haiku 5.5 when using Google Cloud's Agent Platform.
ANTHROPIC_API_KEY, AWS_SECRET_ACCESS_KEY Removes it.
NPM_TOKEN, DB_PASSWORD Removes it, because the name looks like a credential.
DATABASE_URL that contains a password` Removes it, because the value looks like a credential.
PIP_INDEX_URL or NPM_CONFIG_REGISTRY that contains a password` Keeps the URL and cuts the username and password from it.
GITHUB_TOKEN, GH_TOKEN, GH_ENTERPRISE_TOKEN, GITHUB_ENTERPRISE_TOKEN Leaves it in place, so that gh and scripts that call the GitHub API keep working.
HTTP_PROXY, HTTPS_PROXY Leaves it in place, including a username and password in the URL. The sandbox can set these variables…
GIT_CONFIG_COUNT, GIT_CONFIG_KEY_<n>, GIT_CONFIG_VALUE_<n> Leaves it in place, whatever it holds.
:::

Glossary {.unnumbered}

:::glossary Agent : An AI that can take actions, such as reading and editing files or running commands, in a loop until a task is finished, rather than only replying with text.

Agentic loop : The cycle an agent repeats: gather context, plan, act, check the result.

Auto mode : A permission mode where a separate safety classifier reviews Claude's actions instead of asking you about each one.

Bypass permissions : A permission mode (--dangerously-skip-permissions) that skips all checks. Meant only for isolated, disposable environments.

Checkpoint : An automatic snapshot taken before Claude edits files, which you can return to with /rewind.

Classifier : A second AI model that judges whether an action in auto mode matches what you asked for.

CLAUDE.md : A Markdown file of instructions Claude reads at the start of every session.

CLI : Command-line interface: a program you control by typing commands in a terminal.

Context window : The amount of conversation and file content Claude can consider at once.

Current directory : The folder your terminal is "in." Claude Code works on the files in this folder. Also called the working directory.

Diff : A view showing exactly which lines of a file were added, removed, or changed.

Git : A tool that records the history of your files so you can review changes and go back in time.

Hook : A command you configure to run automatically at a set moment, such as after every file edit.

Markdown : A simple text format that uses symbols like # and - for headings and lists.

MCP : Model Context Protocol, an open standard for connecting AI tools to other apps and data.

MCP server : A connector that gives an AI new tools, such as access to Notion, GitHub, or a database.

Model : The specific AI "brain" doing the work, e.g. Opus (most capable), Sonnet (balanced), Haiku (fastest).

PATH : The list of folders your terminal searches when you type a program's name.

Permission mode : The setting that decides which actions Claude can take without asking you.

Plan mode : A permission mode in which Claude researches and proposes a plan but makes no edits until you approve.

Plugin : An installable bundle of skills, subagents, hooks, and MCP servers.

Prompt : The request you type to Claude. Also, the terminal's "ready for input" marker.

Prompt injection : Hidden instructions in a web page, file, or tool result that try to trick an AI into doing something you didn't ask for.

Sandbox : An isolated environment that limits which files and network addresses a command can reach.

Skill : A saved set of instructions, in a SKILL.md file, that Claude can use automatically or that you can run as a /command.

Subagent : A specialist helper with its own instructions and tools that Claude can delegate work to.

Terminal : A text window for typing commands to your computer. Also called a shell, console, or command prompt.

Token : A small chunk of text, roughly three-quarters of a word, used to measure AI usage.

Worktree : A separate working copy of a Git repository, so parallel tasks don't step on each other. :::