Claude Code Commands (100+): The Reference

You're sitting at the terminal, have Claude Code installed, and you're thinking: "Okay, now what?"

There are now well over 100 documented command entries once you count built-ins, skills, workflows, and aliases. Flags and keyboard shortcuts give you further control over your sessions.

This is the current Claude Code command landscape: built-in slash commands, bundled skills and workflows, CLI subcommands, flags, and keyboard shortcuts. Not every command appears for every account because platform, plan, and provider matter. For how Claude Code stacks up against OpenAI Codex, see our Claude Code vs. OpenAI Codex comparison.

TL;DRKey Takeaways

The official reference documents well over 100 command entries. This page separates built-ins, bundled skills, the Deep Research workflow, and retired commands.
Key shortcuts are Ctrl+G or Ctrl+X Ctrl+E for the external editor, Ctrl+O for Transcript Viewer, Option+P or Alt+P for model selection, and Shift+Tab for permission modes.
CLI flags and environment variables control reasoning effort, context size, background agents, cloud sessions, worktrees, MCP, and privacy. --dangerously-skip-permissions belongs only in isolated environments.

The tables cover Claude Code v2.1.293 (as of October 7, 2026). Anthropic documents them in its official command reference, CLI reference, interactive-mode documentation, installation guide, and the official changelog.

Installing and Starting Claude Code

# Recommended native installation on macOS, Linux, and WSL
curl -fsSL https://claude.ai/install.sh | bash

# Native installation in Windows PowerShell
irm https://claude.ai/install.ps1 | iex

# Verify the installation
claude --version

# Start Claude Code
claude                          # Interactive mode
claude "Fix the bug in auth"    # With initial prompt
claude -p "Analyze this code"   # One-time query without session
claude -c                       # Continue last session

Note

Anthropic now recommends the native installer. It updates itself automatically in the background. The npm installation is still supported, but it is no longer the recommended default.

Tip

Always start Claude Code in the correct project folder, not from your Desktop or user directory.

Slash Commands (Built-in)

These commands are your daily tools when working with Claude Code. Type / followed by any letters to filter. Not all commands are visible to every user, some depend on platform, plan, or environment.

Version badges show the earliest published version in which the command appears in Anthropic's official changelog, releases, or published Claude Code package.

Command/add-dir

Parameter<path>

Since versionv2.0.11

DescriptionGive the current session access to an additional working directory.

Command/advisor

Parameter[model|off]

Since versionv2.1.83

DescriptionEnable or disable Advisor, which consults a second model at key moments.

Command/agents

Parameter

Since versionv1.0.58

DescriptionOpen guidance for managing subagents through Claude or .claude/agents/ files.

Command/artifacts

Parameter

Since versionv2.1.208

DescriptionList personal or shared artifacts and attach one to the session.

Command/auto-mode-setup

Parameter

Since versionv2.1.228

DescriptionDraft Auto mode settings from the project and prior sessions.

Command/autocompact

Parameter[auto|<tokens>]

Since versionv2.1.221

DescriptionSet how full the context window can become before automatic compaction.

Command/autofix-pr

Parameter[prompt]

Since versionv2.1.94

DescriptionStart a web session that can watch a pull request and push CI or review fixes.

Command/background

Parameter[prompt]

Since versionv2.1.119

DescriptionDetach the current session as a background agent. Alias: /bg.

Command/branch

Parameter[name]

Since versionv2.1.77

DescriptionCreate a conversation branch and switch to it without losing the original.

Command/btw

Parameter[question]

Since versionv2.0.0

DescriptionAnswer a side question outside the conversation context.

Command/bug

Parameter[report]

Since versionv0.2.9

DescriptionReport a bug or share the session after consent. Alias: /share.

Command/cd

Parameter<path>

Since versionv2.1.169

DescriptionChange working directory while retaining the conversation and prompt cache.

Command/chrome

Parameter

Since versionv2.0.72

DescriptionOpen Claude in Chrome settings.

Command/clear

Parameter[name]

Since versionv0.2.9

DescriptionStart a new conversation without context. Aliases: /reset and /new.

Command/color

Parameter[color|default]

Since versionv2.1.70

DescriptionSet the prompt-bar color.

Command/compact

Parameter[instructions]

Since versionv0.2.47

DescriptionSummarize the conversation to free context.

Command/config

Parameter[key=value ...]

Since versionv0.2.34

DescriptionOpen settings or set values directly. Alias: /settings.

Command/context

Parameter[all]

Since versionv2.1.14

DescriptionShow what occupies the context window, including optimization suggestions.

Command/copy

Parameter[N]

Since versionv0.2.9

DescriptionCopy the latest or Nth-most-recent response to the clipboard.

Command/cost

Parameter

Since versionv1.0.85

DescriptionAlias for /usage.

Command/design-login

Parameter

Since versionv2.1.178

DescriptionAuthorize access to Claude Design for /design-sync.

Command/desktop

Parameter

Since versionv2.0.51

DescriptionContinue the session in the Claude Code desktop app. Alias: /app.

Command/diff

Parameter

Since versionv2.0.0

DescriptionOpen the interactive diff viewer for uncommitted and turn-level changes.

Command/effort

Parameter[level|auto|status|ultracode [on|off]]

Since versionv2.1.76

DescriptionChoose reasoning effort. ultracode on and off, plus keeping the current level, work since v2.1.284.

Command/exit

Parameter

Since versionv0.2.9

DescriptionExit the CLI or detach an attached background session. Alias: /quit.

Command/export

Parameter[filename]

Since versionv0.2.9

DescriptionExport the current conversation as plain text.

Command/fast

Parameter[on|off]

Since versionv2.1.36

DescriptionToggle Fast mode when the account supports it.

Command/feedback

Parameter[report]

Since versionv0.2.9

DescriptionOpen the same feedback dialog as /bug.

Command/focus

Parameter

Since versionv2.1.110

DescriptionSwitch to the reduced focus view in the fullscreen renderer.

Command/fork

Parameter[prompt]

Since versionv2.0.0

DescriptionCopy the conversation into a new background session.

Command/goal

Parameter[condition|clear]

Since versionv2.1.139

DescriptionSet a goal that Claude continues pursuing across turns.

Command/heapdump

Parameter

Since versionv2.1.71

DescriptionWrite diagnostics for memory issues. This hidden command can contain sensitive data.

Command/help

Parameter

Since versionv0.2.9

DescriptionShow help and the commands available in this session.

Command/hooks

Parameter

Since versionv1.0.38

DescriptionShow hook configurations.

Command/ide

Parameter

Since versionv2.0.0

DescriptionManage IDE integrations and display their status.

Command/import

Parameter[codex|gemini|cursor] [--dry-run] [--yes]

Since versionv2.1.213

DescriptionImport configuration from Codex, Gemini CLI, or Cursor, including rules, MCP servers, and skills. Cursor is supported since v2.1.265.

Command/init

Parameter

Since versionv0.2.9

DescriptionCreate project guidance in CLAUDE.md.

Command/insights

Parameter

Since versionv2.1.14

DescriptionGenerate an HTML report about local Claude Code sessions and usage patterns.

Command/install-github-app

Parameter

Since versionv1.0.0

DescriptionSet up the Claude GitHub App for a repository.

Command/install-slack-app

Parameter

Since versionv2.0.51

DescriptionInstall the Claude Slack app through browser sign-in.

Command/keybindings

Parameter

Since versionv2.1.18

DescriptionOpen the file for customizable keyboard shortcuts.

Command/list-agents

Parameter

Since versionv2.1.224

DescriptionList reachable subagents, teammates, and sessions. Alias: /peers.

Command/login

Parameter

Since versionv0.2.96

DescriptionSign in to Anthropic.

Command/logout

Parameter

Since versionv0.2.96

DescriptionSign out of the Anthropic account.

Command/mcp

Parameter[reconnect|enable|disable]

Since versionv0.2.50

DescriptionManage MCP connections and OAuth sign-ins.

Command/memory

Parameter

Since versionv1.0.94

DescriptionEdit CLAUDE.md files and automatic memory.

Command/mobile

Parameter

Since versionv2.0.51

DescriptionShow a QR code for the Claude mobile app. Aliases: /ios and /android.

Command/model

Parameter[model]

Since versionv0.2.9

DescriptionSwitch models and save the choice for new sessions.

Command/output-style

Parameter[style]

Since versionv1.0.81

DescriptionList available output styles or switch to one. Available again since v2.1.269.

Command/passes

Parameter

Since versionv2.0.58

DescriptionShare a free Claude Code week when the account is eligible.

Command/permissions

Parameter

Since versionv2.0.67

DescriptionManage allowed, prompted, and denied tool rules. Alias: /allowed-tools.

Command/plan

Parameter[description]

Since versionv2.1.0

DescriptionEnter plan mode and optionally start with a task.

Command/plugin

Parameter[subcommand]

Since versionv2.0.12

DescriptionManage plugins, for example list, install, enable, and disable.

Command/powerup

Parameter

Since versionv2.1.90

DescriptionOpen interactive mini-lessons about Claude Code features.

Command/privacy-settings

Parameter

Since versionv2.0.51

DescriptionOpen privacy settings for Pro and Max subscribers.

Command/radio

Parameter

Since versionv2.1.132

DescriptionOpen Claude FM in the browser when the command is available for the provider.

Command/rate-limit-options

Parameter

Since versionNot documented

DescriptionShow options after reaching a usage limit. Requires a claude.ai subscription; wait-and-continue options require v2.1.234 or later. Listed in the command menu since v2.1.284; the first version of the previously hidden command is not documented.

Command/recap

Parameter

Since versionv2.1.108

DescriptionGenerate a one-sentence summary of the current session.

Command/release-notes

Parameter

Since versionv2.0.32

DescriptionOpen release notes in the interactive picker.

Command/reload-plugins

Parameter[--force]

Since versionv2.1.69

DescriptionReload active plugins and apply MCP changes with --force.

Command/reload-skills

Parameter

Since versionv2.1.152

DescriptionReload skill and command directories without restarting Claude Code.

Command/remote-control

Parameter

Since versionv2.1.51

DescriptionMake the local session controllable through claude.ai or the Claude app. Alias: /rc.

Command/remote-env

Parameter

Since versionv2.0.47

DescriptionChoose the default environment for cloud agents.

Command/rename

Parameter[name]

Since versionv2.1.20

DescriptionRename the current session.

Command/resume

Parameter[session]

Since versionv1.0.27

DescriptionResume a conversation by ID or name. Directly attach to a running background session since v2.1.285. Alias: /continue.

Command/review

Parameter[level] [--fix] [--comment] [--max-findings <value>] [target]

Since versionv0.2.9

DescriptionAlias for /code-review.

Command/rewind

Parameter

Since versionv2.0.0

DescriptionRestore code and/or conversation to a checkpoint. Aliases: /checkpoint and /undo.

Command/sandbox

Parameter

Since versionv2.0.24

DescriptionToggle sandbox mode on supported platforms.

Command/schedule

Parameter[description]

Since versionv2.1.80

DescriptionCreate, update, or run cloud routines. Alias: /routines.

Command/scroll-speed

Parameter

Since versionv2.1.139

DescriptionSet scrolling speed in the fullscreen renderer.

Command/security-review

Parameter

Since versionv2.1.70

DescriptionCheck the branch diff for security problems.

Command/setup-bedrock

Parameter

Since versionv2.1.92

DescriptionSet up Amazon Bedrock authentication, region, and models.

Command/setup-vertex

Parameter

Since versionv2.1.98

DescriptionSet up Google Cloud Agent Platform.

Command/skills

Parameter

Since versionv2.1.3

DescriptionList available skills and manage their visibility.

Command/stats

Parameter

Since versionv2.0.64

DescriptionAlias for /usage and open the Stats tab.

Command/status

Parameter

Since versionv0.2.9

DescriptionShow version, model, account, connectivity, and session type.

Command/statusline

Parameter

Since versionv1.0.71

DescriptionConfigure the status line.

Command/stickers

Parameter

Since versionv2.0.51

DescriptionOrder Claude Code stickers.

Command/stop

Parameter

Since versionv2.1.119

DescriptionStop the currently attached background session.

Command/subtask

Parameter<task>

Since versionv2.1.212

DescriptionStart a background subagent whose result returns to this conversation.

Command/tasks

Parameter

Since versionv2.0.64

DescriptionShow and manage background work in the current session. Alias: /bashes.

Command/team-onboarding

Parameter

Since versionv2.1.101

DescriptionCreate an onboarding guide for teammates from your usage history.

Command/teleport

Parameter

Since versionv2.1.0

DescriptionBring a cloud session into this terminal. Alias: /tp.

Command/terminal-setup

Parameter

Since versionv1.0.110

DescriptionConfigure terminal-specific shortcuts such as Shift+Enter.

Command/theme

Parameter

Since versionv2.0.74

DescriptionSwitch theme, ANSI theme, or accessible color variant.

Command/tui

Parameter[default|fullscreen]

Since versionv2.1.110

DescriptionSwitch the terminal renderer and restart the conversation in it.

Command/ultrareview

Parameter[PR|branch]

Since versionv2.1.111

DescriptionAlias for the cloud version of /code-review ultra.

Command/upgrade

Parameter

Since versionv2.0.51

DescriptionOpen the upgrade page for a higher plan.

Command/usage

Parameter

Since versionv2.0.0

DescriptionShow costs, plan limits, and usage statistics.

Command/usage-credits

Parameter

Since versionv2.1.144

DescriptionConfigure usage credits or request them from an administrator.

Command/voice

Parameter[hold|tap|off]

Since versionv2.1.69

DescriptionControl voice dictation for Claude.ai accounts.

Command/web-setup

Parameter

Since versionv2.1.79

DescriptionConnect local GitHub credentials to Claude Code on the web.

Command/workflows

Parameter

Since versionv2.1.154

DescriptionOpen progress for running and completed workflows.


`/add-dir`	`<path>`	v2.0.11	Give the current session access to an additional working directory.
`/advisor`	`[model\|off]`	v2.1.83	Enable or disable Advisor, which consults a second model at key moments.
`/agents`		v1.0.58	Open guidance for managing subagents through Claude or .claude/agents/ files.
`/artifacts`		v2.1.208	List personal or shared artifacts and attach one to the session.
`/auto-mode-setup`		v2.1.228	Draft Auto mode settings from the project and prior sessions.
`/autocompact`	`[auto\|<tokens>]`	v2.1.221	Set how full the context window can become before automatic compaction.
`/autofix-pr`	`[prompt]`	v2.1.94	Start a web session that can watch a pull request and push CI or review fixes.
`/background`	`[prompt]`	v2.1.119	Detach the current session as a background agent. Alias: /bg.
`/branch`	`[name]`	v2.1.77	Create a conversation branch and switch to it without losing the original.
`/btw`	`[question]`	v2.0.0	Answer a side question outside the conversation context.
`/bug`	`[report]`	v0.2.9	Report a bug or share the session after consent. Alias: /share.
`/cd`	`<path>`	v2.1.169	Change working directory while retaining the conversation and prompt cache.
`/chrome`		v2.0.72	Open Claude in Chrome settings.
`/clear`	`[name]`	v0.2.9	Start a new conversation without context. Aliases: /reset and /new.
`/color`	`[color\|default]`	v2.1.70	Set the prompt-bar color.
`/compact`	`[instructions]`	v0.2.47	Summarize the conversation to free context.
`/config`	`[key=value ...]`	v0.2.34	Open settings or set values directly. Alias: /settings.
`/context`	`[all]`	v2.1.14	Show what occupies the context window, including optimization suggestions.
`/copy`	`[N]`	v0.2.9	Copy the latest or Nth-most-recent response to the clipboard.
`/cost`		v1.0.85	Alias for /usage.
`/design-login`		v2.1.178	Authorize access to Claude Design for /design-sync.
`/desktop`		v2.0.51	Continue the session in the Claude Code desktop app. Alias: /app.
`/diff`		v2.0.0	Open the interactive diff viewer for uncommitted and turn-level changes.
`/effort`	`[level\|auto\|status\|ultracode [on\|off]]`	v2.1.76	Choose reasoning effort. ultracode on and off, plus keeping the current level, work since v2.1.284.
`/exit`		v0.2.9	Exit the CLI or detach an attached background session. Alias: /quit.
`/export`	`[filename]`	v0.2.9	Export the current conversation as plain text.
`/fast`	`[on\|off]`	v2.1.36	Toggle Fast mode when the account supports it.
`/feedback`	`[report]`	v0.2.9	Open the same feedback dialog as /bug.
`/focus`		v2.1.110	Switch to the reduced focus view in the fullscreen renderer.
`/fork`	`[prompt]`	v2.0.0	Copy the conversation into a new background session.
`/goal`	`[condition\|clear]`	v2.1.139	Set a goal that Claude continues pursuing across turns.
`/heapdump`		v2.1.71	Write diagnostics for memory issues. This hidden command can contain sensitive data.
`/help`		v0.2.9	Show help and the commands available in this session.
`/hooks`		v1.0.38	Show hook configurations.
`/ide`		v2.0.0	Manage IDE integrations and display their status.
`/import`	`[codex\|gemini\|cursor] [--dry-run] [--yes]`	v2.1.213	Import configuration from Codex, Gemini CLI, or Cursor, including rules, MCP servers, and skills. Cursor is supported since v2.1.265.
`/init`		v0.2.9	Create project guidance in CLAUDE.md.
`/insights`		v2.1.14	Generate an HTML report about local Claude Code sessions and usage patterns.
`/install-github-app`		v1.0.0	Set up the Claude GitHub App for a repository.
`/install-slack-app`		v2.0.51	Install the Claude Slack app through browser sign-in.
`/keybindings`		v2.1.18	Open the file for customizable keyboard shortcuts.
`/list-agents`		v2.1.224	List reachable subagents, teammates, and sessions. Alias: /peers.
`/login`		v0.2.96	Sign in to Anthropic.
`/logout`		v0.2.96	Sign out of the Anthropic account.
`/mcp`	`[reconnect\|enable\|disable]`	v0.2.50	Manage MCP connections and OAuth sign-ins.
`/memory`		v1.0.94	Edit CLAUDE.md files and automatic memory.
`/mobile`		v2.0.51	Show a QR code for the Claude mobile app. Aliases: /ios and /android.
`/model`	`[model]`	v0.2.9	Switch models and save the choice for new sessions.
`/output-style`	`[style]`	v1.0.81	List available output styles or switch to one. Available again since v2.1.269.
`/passes`		v2.0.58	Share a free Claude Code week when the account is eligible.
`/permissions`		v2.0.67	Manage allowed, prompted, and denied tool rules. Alias: /allowed-tools.
`/plan`	`[description]`	v2.1.0	Enter plan mode and optionally start with a task.
`/plugin`	`[subcommand]`	v2.0.12	Manage plugins, for example list, install, enable, and disable.
`/powerup`		v2.1.90	Open interactive mini-lessons about Claude Code features.
`/privacy-settings`		v2.0.51	Open privacy settings for Pro and Max subscribers.
`/radio`		v2.1.132	Open Claude FM in the browser when the command is available for the provider.
`/rate-limit-options`		Not documented	Show options after reaching a usage limit. Requires a claude.ai subscription; wait-and-continue options require v2.1.234 or later. Listed in the command menu since v2.1.284; the first version of the previously hidden command is not documented.
`/recap`		v2.1.108	Generate a one-sentence summary of the current session.
`/release-notes`		v2.0.32	Open release notes in the interactive picker.
`/reload-plugins`	`[--force]`	v2.1.69	Reload active plugins and apply MCP changes with --force.
`/reload-skills`		v2.1.152	Reload skill and command directories without restarting Claude Code.
`/remote-control`		v2.1.51	Make the local session controllable through claude.ai or the Claude app. Alias: /rc.
`/remote-env`		v2.0.47	Choose the default environment for cloud agents.
`/rename`	`[name]`	v2.1.20	Rename the current session.
`/resume`	`[session]`	v1.0.27	Resume a conversation by ID or name. Directly attach to a running background session since v2.1.285. Alias: /continue.
`/review`	`[level] [--fix] [--comment] [--max-findings <value>] [target]`	v0.2.9	Alias for /code-review.
`/rewind`		v2.0.0	Restore code and/or conversation to a checkpoint. Aliases: /checkpoint and /undo.
`/sandbox`		v2.0.24	Toggle sandbox mode on supported platforms.
`/schedule`	`[description]`	v2.1.80	Create, update, or run cloud routines. Alias: /routines.
`/scroll-speed`		v2.1.139	Set scrolling speed in the fullscreen renderer.
`/security-review`		v2.1.70	Check the branch diff for security problems.
`/setup-bedrock`		v2.1.92	Set up Amazon Bedrock authentication, region, and models.
`/setup-vertex`		v2.1.98	Set up Google Cloud Agent Platform.
`/skills`		v2.1.3	List available skills and manage their visibility.
`/stats`		v2.0.64	Alias for /usage and open the Stats tab.
`/status`		v0.2.9	Show version, model, account, connectivity, and session type.
`/statusline`		v1.0.71	Configure the status line.
`/stickers`		v2.0.51	Order Claude Code stickers.
`/stop`		v2.1.119	Stop the currently attached background session.
`/subtask`	`<task>`	v2.1.212	Start a background subagent whose result returns to this conversation.
`/tasks`		v2.0.64	Show and manage background work in the current session. Alias: /bashes.
`/team-onboarding`		v2.1.101	Create an onboarding guide for teammates from your usage history.
`/teleport`		v2.1.0	Bring a cloud session into this terminal. Alias: /tp.
`/terminal-setup`		v1.0.110	Configure terminal-specific shortcuts such as Shift+Enter.
`/theme`		v2.0.74	Switch theme, ANSI theme, or accessible color variant.
`/tui`	`[default\|fullscreen]`	v2.1.110	Switch the terminal renderer and restart the conversation in it.
`/ultrareview`	`[PR\|branch]`	v2.1.111	Alias for the cloud version of /code-review ultra.
`/upgrade`		v2.0.51	Open the upgrade page for a higher plan.
`/usage`		v2.0.0	Show costs, plan limits, and usage statistics.
`/usage-credits`		v2.1.144	Configure usage credits or request them from an administrator.
`/voice`	`[hold\|tap\|off]`	v2.1.69	Control voice dictation for Claude.ai accounts.
`/web-setup`		v2.1.79	Connect local GitHub credentials to Claude Code on the web.
`/workflows`		v2.1.154	Open progress for running and completed workflows.

Tip

My favorite command /clear is a real token saver. Almost reached the message limit? This command clears the context and gives you a fresh start.

Bundled Skills and Workflows

Bundled skills are detailed work instructions rather than hard-coded UI logic. Claude can select them itself when the task fits, except for /verify. Workflows go further and distribute background work across several agents.

Note

You can chain up to six skills at the start of a message, for example /simplify /verify check my latest diff.

Skill/artifact-capabilities

DescriptionLoad the reference for published artifact runtime capabilities, such as connector calls and downloads. Availability depends on artifacts and account; the first version is not documented.

Skill/artifact-diagramming

DescriptionLoad guidance for readable diagrams in artifacts. Requires Claude Code v2.1.221 or later.

Skill/batch <prompt>

DescriptionSplit a large change into independent worktrees, each with a background subagent.

Skill

/claude-api [migrate|upgrade|managed-agents-onboard|prompt-audit|cost-optimize|build-eval|hillclimb|preserved-thinking-migration]

DescriptionLoad API and Managed Agents reference material. migrate and managed-agents-onboard predate v2.1.221. Minimum versions: prompt-audit v2.1.221, upgrade v2.1.236, cost-optimize v2.1.247, build-eval and hillclimb v2.1.259, preserved-thinking-migration v2.1.282.

Skill/claude-in-chrome [task]

DescriptionCarry out browser tasks through Claude in Chrome when the integration is enabled. The first version is not documented.

Skill/code-review [level] [--fix] [--comment] [--max-findings <value>] [target]

DescriptionReview a diff, pull request, branch, or path and optionally apply fixes. --max-findings is available since v2.1.288. Alias: /review.

Skill/dataviz [request]

DescriptionHelp design accessible charts, graphics, and dashboards.

Skill/debug [description]

DescriptionEnable debug logging and analyze the current session log.

Skill/design [brief]

DescriptionCreate UI designs as a Claude Design artifact. Requires v2.1.265 or later, available artifacts, and the Design template; Anthropic API only.

Skill/design-sync [prompt]

DescriptionSync a React design system into Claude Design.

Skill/doctor [prompt-audit [path]]

DescriptionCheck installation and configuration, find unnecessary context, and offer changes for review. prompt-audit requires v2.1.283 or later. Alias: /checkup.

Skill/fewer-permission-prompts

DescriptionSuggest a prioritized allowlist for frequent read-only tool calls.

Skill/loop [interval] [prompt]

DescriptionRepeat a prompt while the session remains open. Alias: /proactive.

Skill/plugin-authoring

DescriptionLoad the reference for developing Claude Code mods. Requires Claude Code v2.1.287 or later.

Skill/run

DescriptionStart and operate the app so Claude can observe changes in practice.

Skill/run-skill-generator

DescriptionCreate project-specific foundations for /run and /verify.

Skill/simplify [target]

DescriptionReview changed code for reuse, simplification, efficiency, and abstraction.

Skill/skill-doctor

DescriptionShow skill context costs and usage. Requires Claude Code v2.1.252 or later and feature-flag fetching.

Skill/slides [brief]

DescriptionCreate a presentation as a Claude Slides artifact. Requires v2.1.265 or later, available artifacts, and the Slides template; Anthropic API only.

Skill/update-config [request]

DescriptionChange matching settings.json entries from a natural-language request. The first version is not documented.

Skill/verify

DescriptionBuild, start, and observe the application alongside tests and type checks.

Skill/workflow-authoring

DescriptionLoad the reference for creating and editing dynamic workflows. Requires Claude Code v2.1.248 or later with Dynamic Workflows enabled.


`/artifact-capabilities`	Load the reference for published artifact runtime capabilities, such as connector calls and downloads. Availability depends on artifacts and account; the first version is not documented.
`/artifact-diagramming`	Load guidance for readable diagrams in artifacts. Requires Claude Code v2.1.221 or later.
`/batch <prompt>`	Split a large change into independent worktrees, each with a background subagent.
`/claude-api [migrate\|upgrade\|managed-agents-onboard\|prompt-audit\|cost-optimize\|build-eval\|hillclimb\|preserved-thinking-migration]`	Load API and Managed Agents reference material. migrate and managed-agents-onboard predate v2.1.221. Minimum versions: prompt-audit v2.1.221, upgrade v2.1.236, cost-optimize v2.1.247, build-eval and hillclimb v2.1.259, preserved-thinking-migration v2.1.282.
`/claude-in-chrome [task]`	Carry out browser tasks through Claude in Chrome when the integration is enabled. The first version is not documented.
`/code-review [level] [--fix] [--comment] [--max-findings <value>] [target]`	Review a diff, pull request, branch, or path and optionally apply fixes. --max-findings is available since v2.1.288. Alias: /review.
`/dataviz [request]`	Help design accessible charts, graphics, and dashboards.
`/debug [description]`	Enable debug logging and analyze the current session log.
`/design [brief]`	Create UI designs as a Claude Design artifact. Requires v2.1.265 or later, available artifacts, and the Design template; Anthropic API only.
`/design-sync [prompt]`	Sync a React design system into Claude Design.
`/doctor [prompt-audit [path]]`	Check installation and configuration, find unnecessary context, and offer changes for review. prompt-audit requires v2.1.283 or later. Alias: /checkup.
`/fewer-permission-prompts`	Suggest a prioritized allowlist for frequent read-only tool calls.
`/loop [interval] [prompt]`	Repeat a prompt while the session remains open. Alias: /proactive.
`/plugin-authoring`	Load the reference for developing Claude Code mods. Requires Claude Code v2.1.287 or later.
`/run`	Start and operate the app so Claude can observe changes in practice.
`/run-skill-generator`	Create project-specific foundations for /run and /verify.
`/simplify [target]`	Review changed code for reuse, simplification, efficiency, and abstraction.
`/skill-doctor`	Show skill context costs and usage. Requires Claude Code v2.1.252 or later and feature-flag fetching.
`/slides [brief]`	Create a presentation as a Claude Slides artifact. Requires v2.1.265 or later, available artifacts, and the Slides template; Anthropic API only.
`/update-config [request]`	Change matching settings.json entries from a natural-language request. The first version is not documented.
`/verify`	Build, start, and observe the application alongside tests and type checks.
`/workflow-authoring`	Load the reference for creating and editing dynamic workflows. Requires Claude Code v2.1.248 or later with Dynamic Workflows enabled.

For /code-review and /review, --max-findings requires a value. A number limits the findings, all shows every finding, and default resets the limit.

Bundled Workflow

Workflow/deep-research <question>

DescriptionFan out web research, cross-check sources, and return a sourced report.


`/deep-research <question>`	Fan out web research, cross-check sources, and return a sourced report.

Retired Commands

Command/pr-comments

Status or replacementRemoved in v2.1.91. Ask Claude to inspect pull-request comments instead.

Command/ultraplan

Status or replacementRemoved. The official reference does not publish a separate removal version. Use plan mode or /plan.

Command/vim

Status or replacementRemoved in v2.1.92. Configure editor mode through /config.

Command/extra-usage

Status or replacementReplaced by /usage-credits.


`/pr-comments`	Removed in v2.1.91. Ask Claude to inspect pull-request comments instead.
`/ultraplan`	Removed. The official reference does not publish a separate removal version. Use plan mode or /plan.
`/vim`	Removed in v2.1.92. Configure editor mode through /config.
`/extra-usage`	Replaced by /usage-credits.

Note

MCP servers can also provide their own commands. These appear in the format /mcp__<server>__<prompt> and are dynamically discovered from connected servers.

Keyboard Shortcuts

Keyboard shortcuts make the difference between productive flow and constant looking things up. Here are all available shortcuts for interactive mode.

General Controls

ShortcutCtrl+C

DescriptionInterrupt the active operation. When idle, the first press clears the prompt and the second exits Claude Code.

ContextStandard interrupt

ShortcutCtrl+X Ctrl+K

DescriptionStop all background subagents and disable artifact auto-replies for the rest of the session

ContextPress twice within 3 seconds to confirm

ShortcutCtrl+D

DescriptionShow the exit confirmation on an empty prompt. Press again within 800 ms to exit.

ContextWith text, delete the character to the right

ShortcutCtrl+G / Ctrl+X Ctrl+E

DescriptionOpen prompt in the external editor

ContextFor longer prompts

ShortcutCtrl+L

DescriptionRedraw the terminal screen

ContextKeeps conversation history

ShortcutCtrl+O

DescriptionOpen Transcript Viewer

ContextInspect the conversation transcript

ShortcutCtrl+R

DescriptionReverse search in command history

ContextInteractive history search

ShortcutCtrl+T

DescriptionToggle task list visibility

ContextShow/hide tasks in status area

ShortcutCtrl+V / Cmd+V / Alt+V

DescriptionPaste an image from the clipboard

ContextCtrl+V, iTerm2 Cmd+V, or Windows/WSL Alt+V

ShortcutCtrl+B

DescriptionBackground the current task

ContextTmux users press it twice

ShortcutCtrl+S

DescriptionStash or restore the current prompt

ContextKeep a draft while checking something else

ShortcutCtrl+Z

DescriptionSuspend the process

ContextUnix terminals

ShortcutTab

DescriptionAccept autocomplete or add a comment to a permission response

ContextPrompt and permission dialogs

ShortcutArrow keys / Ctrl+P / Ctrl+N

DescriptionNavigate prompt history

ContextRecall previous inputs

ShortcutLeft / Right arrow

DescriptionCycle dialog tabs

ContextIn permission dialogs and menus

ShortcutEsc

DescriptionInterrupt generation or close a dialog

ContextSingle Escape

ShortcutEsc Esc

DescriptionClear a draft or open rewind

ContextDepends on the current state

ShortcutOption+T / Alt+T

DescriptionToggle extended thinking

ContextHas no effect with Fable 5, Fable 5.1, Opus 5.5, or Sonnet 5.5, where thinking is always enabled

ShortcutOption+P / Alt+P

DescriptionOpen model picker

ContextQuick model switch without clearing prompt

ShortcutOption+O / Alt+O

DescriptionToggle Fast mode

ContextWhen the account supports it

ShortcutShift+Tab

DescriptionSwitch permission modes

ContextManual, acceptEdits, Plan, bypass, and Auto. Alt+M is the Windows fallback without VT mode.


`Ctrl+C`	Interrupt the active operation. When idle, the first press clears the prompt and the second exits Claude Code.	Standard interrupt
`Ctrl+X Ctrl+K`	Stop all background subagents and disable artifact auto-replies for the rest of the session	Press twice within 3 seconds to confirm
`Ctrl+D`	Show the exit confirmation on an empty prompt. Press again within 800 ms to exit.	With text, delete the character to the right
`Ctrl+G / Ctrl+X Ctrl+E`	Open prompt in the external editor	For longer prompts
`Ctrl+L`	Redraw the terminal screen	Keeps conversation history
`Ctrl+O`	Open Transcript Viewer	Inspect the conversation transcript
`Ctrl+R`	Reverse search in command history	Interactive history search
`Ctrl+T`	Toggle task list visibility	Show/hide tasks in status area
`Ctrl+V / Cmd+V / Alt+V`	Paste an image from the clipboard	Ctrl+V, iTerm2 Cmd+V, or Windows/WSL Alt+V
`Ctrl+B`	Background the current task	Tmux users press it twice
`Ctrl+S`	Stash or restore the current prompt	Keep a draft while checking something else
`Ctrl+Z`	Suspend the process	Unix terminals
`Tab`	Accept autocomplete or add a comment to a permission response	Prompt and permission dialogs
`Arrow keys / Ctrl+P / Ctrl+N`	Navigate prompt history	Recall previous inputs
`Left / Right arrow`	Cycle dialog tabs	In permission dialogs and menus
`Esc`	Interrupt generation or close a dialog	Single Escape
`Esc Esc`	Clear a draft or open rewind	Depends on the current state
`Option+T / Alt+T`	Toggle extended thinking	Has no effect with Fable 5, Fable 5.1, Opus 5.5, or Sonnet 5.5, where thinking is always enabled
`Option+P / Alt+P`	Open model picker	Quick model switch without clearing prompt
`Option+O / Alt+O`	Toggle Fast mode	When the account supports it
`Shift+Tab`	Switch permission modes	Manual, acceptEdits, Plan, bypass, and Auto. Alt+M is the Windows fallback without VT mode.

Text Editing

ShortcutCtrl+A / Ctrl+E

DescriptionMove to the start or end of the current line

ContextLine navigation

ShortcutCtrl+K

DescriptionDelete to end of line

ContextDeleted text is stored

ShortcutCtrl+U

DescriptionDelete from cursor to start of line

ContextDeleted text is stored

ShortcutCtrl+W

DescriptionDelete back to the previous whitespace, including a full path or --flag=value

ContextFor the previous word, use Option+Delete on macOS or Ctrl+Backspace on Windows

ShortcutCtrl+Y

DescriptionPaste previously deleted text

ContextPaste from kill ring

ShortcutAlt+Y (after Ctrl+Y)

DescriptionCycle through paste history

ContextCycle previously deleted texts

ShortcutAlt+B

DescriptionMove cursor back one word

ContextOption as Meta on macOS

ShortcutAlt+F

DescriptionMove cursor forward one word

ContextOption as Meta on macOS

ShortcutAlt+D

DescriptionDelete the next word

ContextOption as Meta on macOS

ShortcutCtrl+_ / Ctrl+Shift+-

DescriptionUndo the last edit

ContextRestores text and cursor position


`Ctrl+A / Ctrl+E`	Move to the start or end of the current line	Line navigation
`Ctrl+K`	Delete to end of line	Deleted text is stored
`Ctrl+U`	Delete from cursor to start of line	Deleted text is stored
`Ctrl+W`	Delete back to the previous whitespace, including a full path or --flag=value	For the previous word, use Option+Delete on macOS or Ctrl+Backspace on Windows
`Ctrl+Y`	Paste previously deleted text	Paste from kill ring
`Alt+Y (after Ctrl+Y)`	Cycle through paste history	Cycle previously deleted texts
`Alt+B`	Move cursor back one word	Option as Meta on macOS
`Alt+F`	Move cursor forward one word	Option as Meta on macOS
`Alt+D`	Delete the next word	Option as Meta on macOS
`Ctrl+_ / Ctrl+Shift+-`	Undo the last edit	Restores text and cursor position

Note

Since v2.1.261, word shortcuts follow Bash Readline conventions. Punctuation separates words, while Ctrl+W deletes back to the previous whitespace. The earlier keybindingFlavor setting no longer has any effect.

Multi-line Input

MethodQuick Escape

Shortcut\\ + Enter

ContextWorks in all terminals

MethodOption key

ShortcutOption+Enter

ContextAfter enabling Option as Meta on macOS

MethodShift+Enter

ShortcutShift+Enter

ContextNative in iTerm2, WezTerm, Ghostty, Kitty, Warp, Apple Terminal, and Windows Terminal

MethodControl Sequence

ShortcutCtrl+J

ContextLine feed for multi-line

MethodPaste Mode

ShortcutDirect paste

ContextFor code blocks, logs


Quick Escape	`\\ + Enter`	Works in all terminals
Option key	`Option+Enter`	After enabling Option as Meta on macOS
Shift+Enter	`Shift+Enter`	Native in iTerm2, WezTerm, Ghostty, Kitty, Warp, Apple Terminal, and Windows Terminal
Control Sequence	`Ctrl+J`	Line feed for multi-line
Paste Mode	`Direct paste`	For code blocks, logs

Quick Commands

Shortcut/ at start

DescriptionExecute slash command or skill

NotesFilters and autocompletes

Shortcut! at start

DescriptionBash mode

NotesExecutes commands directly, output in session

Shortcut@

DescriptionMention a file path or reachable session

NotesTriggers path autocomplete

Shortcut:

DescriptionInsert an emoji shortcut

NotesSuggestions appear after at least two characters

Shortcut? on an empty prompt

DescriptionToggle shortcut help

NotesWith text in the prompt, it inserts a question mark


`/ at start`	Execute slash command or skill	Filters and autocompletes
`! at start`	Bash mode	Executes commands directly, output in session
`@`	Mention a file path or reachable session	Triggers path autocomplete
`:`	Insert an emoji shortcut	Suggestions appear after at least two characters
`?` on an empty prompt	Toggle shortcut help	With text in the prompt, it inserts a question mark

Theme and Voice Input

ShortcutCtrl+T in /theme

DescriptionToggle syntax highlighting for code blocks

ContextOnly inside the theme picker

ShortcutHold or tap Space

DescriptionStart or stop voice dictation

ContextAfter enabling voice input. Run /voice tap first for tap-to-toggle.


`Ctrl+T` in `/theme`	Toggle syntax highlighting for code blocks	Only inside the theme picker
`Hold or tap Space`	Start or stop voice dictation	After enabling voice input. Run /voice tap first for tap-to-toggle.

Transcript Viewer

Ctrl+O opens the full conversation with tool calls and timestamps. The viewer has its own shortcuts.

Shortcut?

DescriptionToggle Transcript Viewer help in the fullscreen renderer

Shortcut{ / }

DescriptionJump to the previous or next user prompt in the fullscreen renderer

ShortcutCtrl+E

DescriptionToggle all content in the classic renderer

Shortcut[

DescriptionWrite the full conversation from fullscreen mode to native terminal scrollback

Shortcutv

DescriptionWrite the conversation to a temporary file and open it in $VISUAL or $EDITOR

Shortcutq / Ctrl+C / Esc

DescriptionClose the Transcript Viewer


`?`	Toggle Transcript Viewer help in the fullscreen renderer
`{ / }`	Jump to the previous or next user prompt in the fullscreen renderer
`Ctrl+E`	Toggle all content in the classic renderer
`[`	Write the full conversation from fullscreen mode to native terminal scrollback
`v`	Write the conversation to a temporary file and open it in $VISUAL or $EDITOR
`q / Ctrl+C / Esc`	Close the Transcript Viewer

CLI Commands

For everyday work, starting, resuming, and claude mcp are usually enough. The CLI can now also manage background sessions, cloud environments, plugins, and local project data.

Commandclaude

DescriptionStart interactive REPL

Exampleclaude

Commandclaude "query"

DescriptionStart REPL with initial prompt

Exampleclaude "explain this project"

Commandclaude -p "query"

DescriptionQuery via SDK, then exit

Exampleclaude -p "explain this function"

Commandcat file | claude -p "query"

DescriptionProcess piped content

Examplecat logs.txt | claude -p "explain"

Commandclaude -c

DescriptionContinue most recent conversation

Exampleclaude -c

Commandclaude -c -p "query"

DescriptionContinue via SDK

Exampleclaude -c -p "Check for type errors"

Commandclaude -r "session" "query"

DescriptionResume session by ID or name

Exampleclaude -r "auth-refactor" "Finish this PR"

Commandclaude update

DescriptionUpdate to latest version

Exampleclaude update

Commandclaude auth login

DescriptionSign in to Anthropic account (with --email or --sso)

Exampleclaude auth login --sso

Commandclaude auth logout

DescriptionSign out from Anthropic account

Exampleclaude auth logout

Commandclaude auth status

DescriptionShow authentication status (JSON or --text)

Exampleclaude auth status --text

Commandclaude agents

DescriptionOpen the agent view, optionally as JSON or for a specific directory

Exampleclaude agents --json

Commandclaude doctor

DescriptionCheck Claude Code installation health

Exampleclaude doctor

Commandclaude mcp

DescriptionConfigure and manage MCP servers. Add remote HTTP servers with --transport http.

Exampleclaude mcp add --transport http github https://api.githubcopilot.com/mcp/

Commandclaude mcp login | logout

DescriptionSign in to or out of an MCP server

Exampleclaude mcp login github

Commandclaude gateway --config

DescriptionStart an API gateway with a configuration file

Exampleclaude gateway --config gateway.yaml

Commandclaude install [version]

DescriptionInstall a specific Claude Code version

Exampleclaude install latest

Commandclaude attach <id|name>

DescriptionAttach the terminal to a background session. Partial matching of a running session name is available since v2.1.290.

Exampleclaude attach 7c5dcf5d

Commandclaude auto-mode defaults | config | reset

DescriptionInspect defaults, effective configuration, or reset Auto mode

Exampleclaude auto-mode config

Commandclaude daemon logs|run|status

DescriptionFollow supervisor logs, run it in the foreground, or inspect its status.

Exampleclaude daemon status

Commandclaude daemon stop --any [--keep-workers]

DescriptionStop the supervisor across projects. Add --keep-workers to leave its worker sessions running.

Exampleclaude daemon stop --any

Commandclaude import [codex|gemini|cursor]

DescriptionImport configuration from another coding agent. Cursor import is available since v2.1.265.

Exampleclaude import codex --dry-run

Commandclaude logs <id|name>

DescriptionShow recent output from a background session. Partial matching of a running session name is available since v2.1.290.

Exampleclaude logs 7c5dcf5d

Commandclaude plugin

DescriptionManage plugins. The plural form plugins is an alias. Install with --marketplace <source>, available since v2.1.292.

Exampleclaude plugin install code-review --marketplace anthropics/claude-plugins-official

Commandclaude purge [path]

DescriptionDelete local transcripts, task lists, debug logs, edit and prompt history, and the project entry. Preview with --dry-run first. The previous name claude project purge still works and prints a notice.

Exampleclaude purge ~/work/repo --dry-run

Commandclaude remote-control

DescriptionStart Remote Control so a session can be controlled via claude.ai

Exampleclaude remote-control --name "My Project"

Commandclaude respawn | rm | stop | kill

DescriptionRestart, remove, or stop managed sessions. kill is an alias for stop.

Exampleclaude stop

Commandclaude self-hosted-runner

DescriptionManage a self-hosted runner

Exampleclaude self-hosted-runner

Commandclaude setup-token

DescriptionCreate or manage a setup token

Exampleclaude setup-token

Commandclaude ultrareview

DescriptionRun the cloud code-review variant

Exampleclaude ultrareview


`claude`	Start interactive REPL	`claude`
`claude "query"`	Start REPL with initial prompt	`claude "explain this project"`
`claude -p "query"`	Query via SDK, then exit	`claude -p "explain this function"`
`cat file \| claude -p "query"`	Process piped content	`cat logs.txt \| claude -p "explain"`
`claude -c`	Continue most recent conversation	`claude -c`
`claude -c -p "query"`	Continue via SDK	`claude -c -p "Check for type errors"`
`claude -r "session" "query"`	Resume session by ID or name	`claude -r "auth-refactor" "Finish this PR"`
`claude update`	Update to latest version	`claude update`
`claude auth login`	Sign in to Anthropic account (with --email or --sso)	`claude auth login --sso`
`claude auth logout`	Sign out from Anthropic account	`claude auth logout`
`claude auth status`	Show authentication status (JSON or --text)	`claude auth status --text`
`claude agents`	Open the agent view, optionally as JSON or for a specific directory	`claude agents --json`
`claude doctor`	Check Claude Code installation health	`claude doctor`
`claude mcp`	Configure and manage MCP servers. Add remote HTTP servers with --transport http.	`claude mcp add --transport http github https://api.githubcopilot.com/mcp/`
`claude mcp login \| logout`	Sign in to or out of an MCP server	`claude mcp login github`
`claude gateway --config`	Start an API gateway with a configuration file	`claude gateway --config gateway.yaml`
`claude install [version]`	Install a specific Claude Code version	`claude install latest`
`claude attach <id\|name>`	Attach the terminal to a background session. Partial matching of a running session name is available since v2.1.290.	`claude attach 7c5dcf5d`
`claude auto-mode defaults \| config \| reset`	Inspect defaults, effective configuration, or reset Auto mode	`claude auto-mode config`
`claude daemon logs\|run\|status`	Follow supervisor logs, run it in the foreground, or inspect its status.	`claude daemon status`
`claude daemon stop --any [--keep-workers]`	Stop the supervisor across projects. Add --keep-workers to leave its worker sessions running.	`claude daemon stop --any`
`claude import [codex\|gemini\|cursor]`	Import configuration from another coding agent. Cursor import is available since v2.1.265.	`claude import codex --dry-run`
`claude logs <id\|name>`	Show recent output from a background session. Partial matching of a running session name is available since v2.1.290.	`claude logs 7c5dcf5d`
`claude plugin`	Manage plugins. The plural form plugins is an alias. Install with --marketplace <source>, available since v2.1.292.	`claude plugin install code-review --marketplace anthropics/claude-plugins-official`
`claude purge [path]`	Delete local transcripts, task lists, debug logs, edit and prompt history, and the project entry. Preview with --dry-run first. The previous name claude project purge still works and prints a notice.	`claude purge ~/work/repo --dry-run`
`claude remote-control`	Start Remote Control so a session can be controlled via claude.ai	`claude remote-control --name "My Project"`
`claude respawn \| rm \| stop \| kill`	Restart, remove, or stop managed sessions. kill is an alias for stop.	`claude stop`
`claude self-hosted-runner`	Manage a self-hosted runner	`claude self-hosted-runner`
`claude setup-token`	Create or manage a setup token	`claude setup-token`
`claude ultrareview`	Run the cloud code-review variant	`claude ultrareview`

CLI Flags

CLI flags are very useful for starting Claude Code with the right settings.

Basic Flags

Flag-p, --print

DescriptionPrint response without interactive mode

Exampleclaude -p "query"

Flag-c, --continue

DescriptionContinue last conversation

Exampleclaude -c

Flag-r, --resume

DescriptionResume specific session by ID or name

Exampleclaude -r "auth-refactor" "query"

Flag-n, --name

DescriptionSet display name for session (visible in /resume and terminal title)

Exampleclaude -n "feature-work"

Flag-w, --worktree

DescriptionStart in isolated git worktree

Exampleclaude -w feature-auth

Flag-v, --version

DescriptionOutput version number

Exampleclaude -v


`-p, --print`	Print response without interactive mode	`claude -p "query"`
`-c, --continue`	Continue last conversation	`claude -c`
`-r, --resume`	Resume specific session by ID or name	`claude -r "auth-refactor" "query"`
`-n, --name`	Set display name for session (visible in /resume and terminal title)	`claude -n "feature-work"`
`-w, --worktree`	Start in isolated git worktree	`claude -w feature-auth`
`-v, --version`	Output version number	`claude -v`

Model & System Prompt

Flag--effort

DescriptionSet effort to low, medium, high, xhigh, max, ultracode, or auto when the selected model supports it

Exampleclaude --effort high

Flag--model

DescriptionSet a model with an alias such as sonnet, opus, haiku, or fable, or a full model ID. Claude Haiku 5.5 (claude-haiku-5-5) is the default Haiku model on the Anthropic API.

Exampleclaude --model sonnet

Flag--system-prompt

DescriptionReplace entire system prompt

Exampleclaude --system-prompt "You are a Python expert"

Flag--system-prompt-file

DescriptionLoad system prompt from file (replaces default)

Exampleclaude --system-prompt-file ./prompt.txt

Flag--append-system-prompt

DescriptionAppend text to default system prompt

Exampleclaude --append-system-prompt "Always use TypeScript"

Flag--append-system-prompt-file

DescriptionAppend file contents to default system prompt

Exampleclaude --append-system-prompt-file ./rules.txt

Flag--system-prompt-snapshot

DescriptionUse off to rebuild the system prompt on each request. Available since v2.1.257.

Exampleclaude --system-prompt-snapshot off

Flag--exclude-dynamic-system-prompt-sections

DescriptionMove machine-specific prompt sections into the first user message to improve prompt-cache reuse for scripts

Exampleclaude -p --exclude-dynamic-system-prompt-sections "query"

Flag--tools

DescriptionRestrict available tools ("" for none, "default" for all)

Exampleclaude --tools "Bash,Edit,Read"


`--effort`	Set effort to low, medium, high, xhigh, max, ultracode, or auto when the selected model supports it	`claude --effort high`
`--model`	Set a model with an alias such as sonnet, opus, haiku, or fable, or a full model ID. Claude Haiku 5.5 (claude-haiku-5-5) is the default Haiku model on the Anthropic API.	`claude --model sonnet`
`--system-prompt`	Replace entire system prompt	`claude --system-prompt "You are a Python expert"`
`--system-prompt-file`	Load system prompt from file (replaces default)	`claude --system-prompt-file ./prompt.txt`
`--append-system-prompt`	Append text to default system prompt	`claude --append-system-prompt "Always use TypeScript"`
`--append-system-prompt-file`	Append file contents to default system prompt	`claude --append-system-prompt-file ./rules.txt`
`--system-prompt-snapshot`	Use off to rebuild the system prompt on each request. Available since v2.1.257.	`claude --system-prompt-snapshot off`
`--exclude-dynamic-system-prompt-sections`	Move machine-specific prompt sections into the first user message to improve prompt-cache reuse for scripts	`claude -p --exclude-dynamic-system-prompt-sections "query"`
`--tools`	Restrict available tools ("" for none, "default" for all)	`claude --tools "Bash,Edit,Read"`

Agents & Subagents

Flag--agent

DescriptionSpecify an agent for the session

Exampleclaude --agent my-custom-agent

Flag--agents

DescriptionDefine subagents via JSON. In print mode, a JSON file path also works, since v2.1.281.

Exampleclaude -p --agents ./agents.json "query"

Flag--append-subagent-system-prompt

DescriptionAppend text to every subagent system prompt in print mode

Exampleclaude -p --append-subagent-system-prompt "Cite file paths" "query"

Flag--append-subagent-system-prompt-file

DescriptionAppend file contents to subagent system prompts in print mode. Available since v2.1.261; cannot be combined with the text flag.

Exampleclaude -p --append-subagent-system-prompt-file ./subagent-rules.txt "query"

Flag--teammate-mode

DescriptionSet Agent Teams display mode: in-process, auto, tmux, or iterm2

Exampleclaude --teammate-mode auto


`--agent`	Specify an agent for the session	`claude --agent my-custom-agent`
`--agents`	Define subagents via JSON. In print mode, a JSON file path also works, since v2.1.281.	`claude -p --agents ./agents.json "query"`
`--append-subagent-system-prompt`	Append text to every subagent system prompt in print mode	`claude -p --append-subagent-system-prompt "Cite file paths" "query"`
`--append-subagent-system-prompt-file`	Append file contents to subagent system prompts in print mode. Available since v2.1.261; cannot be combined with the text flag.	`claude -p --append-subagent-system-prompt-file ./subagent-rules.txt "query"`
`--teammate-mode`	Set Agent Teams display mode: in-process, auto, tmux, or iterm2	`claude --teammate-mode auto`

Permissions & Security

Flag--allowedTools / --allowed-tools

DescriptionAllow tools without permission prompt

Exampleclaude --allowedTools "Bash(git log *)" "Read"

Flag--disallowedTools / --disallowed-tools

DescriptionRemove tools entirely or deny matching calls

Exampleclaude --disallowedTools "Edit" "Bash(curl *)"

Flag--permission-mode

DescriptionStart in specific permission mode

Exampleclaude --permission-mode plan

Flag--allow-dangerously-skip-permissions

DescriptionAdd bypass mode to Shift+Tab without starting the session in it

Exampleclaude --permission-mode plan --allow-dangerously-skip-permissions

Flag--dangerously-skip-permissions

DescriptionSkip all permission prompts (use with caution!)

Exampleclaude --dangerously-skip-permissions

Flag--permission-prompt-tool

DescriptionMCP tool for permission prompts (non-interactive)

Exampleclaude -p --permission-prompt-tool mcp_tool "query"

Flag--permission-prompts

DescriptionChoose who answers permission prompts in print mode. none denies them automatically. Available since v2.1.259.

Exampleclaude -p --permission-prompts none "query"


`--allowedTools / --allowed-tools`	Allow tools without permission prompt	`claude --allowedTools "Bash(git log *)" "Read"`
`--disallowedTools / --disallowed-tools`	Remove tools entirely or deny matching calls	`claude --disallowedTools "Edit" "Bash(curl *)"`
`--permission-mode`	Start in specific permission mode	`claude --permission-mode plan`
`--allow-dangerously-skip-permissions`	Add bypass mode to Shift+Tab without starting the session in it	`claude --permission-mode plan --allow-dangerously-skip-permissions`
`--dangerously-skip-permissions`	Skip all permission prompts (use with caution!)	`claude --dangerously-skip-permissions`
`--permission-prompt-tool`	MCP tool for permission prompts (non-interactive)	`claude -p --permission-prompt-tool mcp_tool "query"`
`--permission-prompts`	Choose who answers permission prompts in print mode. none denies them automatically. Available since v2.1.259.	`claude -p --permission-prompts none "query"`

MCP Server & Integration

Flag--mcp-config

DescriptionLoad MCP servers from JSON files (space-separated)

Exampleclaude --mcp-config ./mcp.json

Flag--strict-mcp-config

DescriptionUse ONLY --mcp-config, ignore all other MCP sources

Exampleclaude --strict-mcp-config --mcp-config ./mcp.json

Flag--plugin-dir

DescriptionLoad plugins from a directory (repeat flag for multiple)

Exampleclaude --plugin-dir ./my-plugins

Flag--plugin-url

DescriptionLoad a plugin ZIP from a URL for this session

Exampleclaude --plugin-url https://example.com/plugin.zip

Flag--chrome / --no-chrome

DescriptionEnable/disable Chrome browser integration

Exampleclaude --chrome

Flag--ide

DescriptionAutomatically connect to IDE at startup

Exampleclaude --ide

Flag--channels

DescriptionResearch Preview for MCP channel notifications from plugins

Exampleclaude --channels plugin:my-notifier@my-marketplace

Flag--dangerously-load-development-channels

DescriptionLoad unapproved channels for local development after confirmation

Exampleclaude --dangerously-load-development-channels server:webhook

Flag--betas

DescriptionBeta headers for API requests (API key only)

Exampleclaude --betas "feature-x"


`--mcp-config`	Load MCP servers from JSON files (space-separated)	`claude --mcp-config ./mcp.json`
`--strict-mcp-config`	Use ONLY --mcp-config, ignore all other MCP sources	`claude --strict-mcp-config --mcp-config ./mcp.json`
`--plugin-dir`	Load plugins from a directory (repeat flag for multiple)	`claude --plugin-dir ./my-plugins`
`--plugin-url`	Load a plugin ZIP from a URL for this session	`claude --plugin-url https://example.com/plugin.zip`
`--chrome / --no-chrome`	Enable/disable Chrome browser integration	`claude --chrome`
`--ide`	Automatically connect to IDE at startup	`claude --ide`
`--channels`	Research Preview for MCP channel notifications from plugins	`claude --channels plugin:my-notifier@my-marketplace`
`--dangerously-load-development-channels`	Load unapproved channels for local development after confirmation	`claude --dangerously-load-development-channels server:webhook`
`--betas`	Beta headers for API requests (API key only)	`claude --betas "feature-x"`

Output & Format

Flag--output-format

DescriptionSpecify output format (text, json, stream-json)

Exampleclaude -p "query" --output-format json

Flag--input-format

DescriptionSpecify input format (text, stream-json)

Exampleclaude -p --input-format stream-json

Flag--json-schema

DescriptionGet validated JSON matching schema (print mode only)

Exampleclaude -p --json-schema '{...}' "query"

Flag--include-partial-messages

DescriptionInclude partial streaming events (stream-json only)

Exampleclaude -p --output-format stream-json --include-partial-messages "query"

Flag--forward-subagent-text

DescriptionEmit subagent text and thinking blocks in stream-json output

Exampleclaude -p --output-format stream-json --verbose --forward-subagent-text "query"

Flag--include-hook-events

DescriptionEmit hook lifecycle events in stream-json output

Exampleclaude -p --output-format stream-json --verbose --include-hook-events "query"

Flag--prompt-suggestions

DescriptionEmit predicted follow-up prompts in stream-json output

Exampleclaude -p --prompt-suggestions --output-format stream-json --verbose "query"

Flag--replay-user-messages

DescriptionRe-emit incoming user messages when both input and output use stream-json

Exampleclaude -p --input-format stream-json --output-format stream-json --verbose --replay-user-messages

Flag--verbose

DescriptionEnable verbose logging with full turn-by-turn output

Exampleclaude --verbose


`--output-format`	Specify output format (text, json, stream-json)	`claude -p "query" --output-format json`
`--input-format`	Specify input format (text, stream-json)	`claude -p --input-format stream-json`
`--json-schema`	Get validated JSON matching schema (print mode only)	`claude -p --json-schema '{...}' "query"`
`--include-partial-messages`	Include partial streaming events (stream-json only)	`claude -p --output-format stream-json --include-partial-messages "query"`
`--forward-subagent-text`	Emit subagent text and thinking blocks in stream-json output	`claude -p --output-format stream-json --verbose --forward-subagent-text "query"`
`--include-hook-events`	Emit hook lifecycle events in stream-json output	`claude -p --output-format stream-json --verbose --include-hook-events "query"`
`--prompt-suggestions`	Emit predicted follow-up prompts in stream-json output	`claude -p --prompt-suggestions --output-format stream-json --verbose "query"`
`--replay-user-messages`	Re-emit incoming user messages when both input and output use stream-json	`claude -p --input-format stream-json --output-format stream-json --verbose --replay-user-messages`
`--verbose`	Enable verbose logging with full turn-by-turn output	`claude --verbose`

Directories & Session

Flag--add-dir

DescriptionAdd additional working directories

Exampleclaude --add-dir ../apps ../lib

Flag--environment

DescriptionStart a cloud session in a self-hosted environment by ID

Exampleclaude -p "Fix the login" --environment ccpool_abc123

Flag--ref

DescriptionUse a named Git ref instead of local HEAD with --environment

Exampleclaude -p "Run the build" --environment ccpool_abc123 --ref main

Flag--from-pr

DescriptionResume sessions linked to a GitHub PR

Exampleclaude --from-pr 123

Flag--session-id

DescriptionUse specific session ID (UUID)

Exampleclaude --session-id "550e8400-..."

Flag--fork-session

DescriptionCreate new session ID when resuming

Exampleclaude --resume abc123 --fork-session

Flag--no-session-persistence

DescriptionDisable session saving (print mode only)

Exampleclaude -p --no-session-persistence "query"

Flag--tmux

DescriptionCreate a tmux session for a worktree, optionally with --tmux=classic

Exampleclaude -w feature-auth --tmux


`--add-dir`	Add additional working directories	`claude --add-dir ../apps ../lib`
`--environment`	Start a cloud session in a self-hosted environment by ID	`claude -p "Fix the login" --environment ccpool_abc123`
`--ref`	Use a named Git ref instead of local HEAD with --environment	`claude -p "Run the build" --environment ccpool_abc123 --ref main`
`--from-pr`	Resume sessions linked to a GitHub PR	`claude --from-pr 123`
`--session-id`	Use specific session ID (UUID)	`claude --session-id "550e8400-..."`
`--fork-session`	Create new session ID when resuming	`claude --resume abc123 --fork-session`
`--no-session-persistence`	Disable session saving (print mode only)	`claude -p --no-session-persistence "query"`
`--tmux`	Create a tmux session for a worktree, optionally with --tmux=classic	`claude -w feature-auth --tmux`

Remote & More

Flag--desktop

DescriptionOpen the Claude Desktop app in the current directory and exit the CLI. Available since v2.1.285; requires macOS or Windows x64 and a Claude subscription.

Exampleclaude --desktop

Flag--cloud

DescriptionStart or target a cloud session on claude.ai

Exampleclaude --cloud "Fix all bugs"

Flag--remote

DescriptionDeprecated alias for --cloud

Exampleclaude --remote "Fix all bugs"

Flag--bg, --background

DescriptionStart a background agent and return its session ID immediately. Do not combine with -p.

Exampleclaude --bg "check the flaky test"

Flag--exec

DescriptionRun a shell command as a PTY-backed background job together with --bg

Exampleclaude --bg --exec 'pytest -x'

Flag--remote-control, --rc

DescriptionStart with Remote Control (session controllable via claude.ai)

Exampleclaude --rc "My Project"

Flag--remote-control-session-name-prefix

DescriptionSet the prefix for automatically generated Remote Control session names

Exampleclaude remote-control --remote-control-session-name-prefix dev-box

Flag--teleport

DescriptionResume remote web session locally

Exampleclaude --teleport

Flag--debug

DescriptionEnable debug logging. Category filters must follow an equals sign.

Exampleclaude --debug='api,mcp'

Flag--debug-file

DescriptionWrite debug logs to a specified file and enable debugging

Exampleclaude --debug-file /tmp/claude-debug.log

Flag--max-turns

DescriptionLimit agentic turns (print mode only)

Exampleclaude -p --max-turns 3 "query"

Flag--max-budget-usd

DescriptionMaximum budget in USD (print mode only)

Exampleclaude -p --max-budget-usd 5.00 "query"

Flag--fallback-model

DescriptionAuto-fallback on overload (print mode only)

Exampleclaude -p --fallback-model sonnet "query"

Flag--init / --maintenance / --init-only

DescriptionRun Setup hooks for initialization or maintenance, or exit after setup

Exampleclaude -p --maintenance "query"

Flag--settings

DescriptionLoad settings from JSON file or string

Exampleclaude --settings ./settings.json

Flag--setting-sources

DescriptionComma-separated list of settings sources (user, project, local)

Exampleclaude --setting-sources "user,project"

Flag--safe-mode

DescriptionTroubleshoot without local customization such as CLAUDE.md, skills, plugins, and hooks

Exampleclaude --safe-mode

Flag--bare

DescriptionMinimal mode for scriptable calls without automatic project customization

Exampleclaude --bare -p "query"

Flag--autocompact

DescriptionSet the auto-compaction window for this session only

Exampleclaude --autocompact 500k

Flag--advisor

DescriptionEnable the server-side Advisor for this session

Exampleclaude --advisor opus

Flag--ax-screen-reader

DescriptionForce simplified, screen-reader-friendly output

Exampleclaude --ax-screen-reader

Flag--disable-slash-commands

DescriptionDisable all skills and commands for this session

Exampleclaude --disable-slash-commands


`--desktop`	Open the Claude Desktop app in the current directory and exit the CLI. Available since v2.1.285; requires macOS or Windows x64 and a Claude subscription.	`claude --desktop`
`--cloud`	Start or target a cloud session on claude.ai	`claude --cloud "Fix all bugs"`
`--remote`	Deprecated alias for --cloud	`claude --remote "Fix all bugs"`
`--bg, --background`	Start a background agent and return its session ID immediately. Do not combine with -p.	`claude --bg "check the flaky test"`
`--exec`	Run a shell command as a PTY-backed background job together with --bg	`claude --bg --exec 'pytest -x'`
`--remote-control, --rc`	Start with Remote Control (session controllable via claude.ai)	`claude --rc "My Project"`
`--remote-control-session-name-prefix`	Set the prefix for automatically generated Remote Control session names	`claude remote-control --remote-control-session-name-prefix dev-box`
`--teleport`	Resume remote web session locally	`claude --teleport`
`--debug`	Enable debug logging. Category filters must follow an equals sign.	`claude --debug='api,mcp'`
`--debug-file`	Write debug logs to a specified file and enable debugging	`claude --debug-file /tmp/claude-debug.log`
`--max-turns`	Limit agentic turns (print mode only)	`claude -p --max-turns 3 "query"`
`--max-budget-usd`	Maximum budget in USD (print mode only)	`claude -p --max-budget-usd 5.00 "query"`
`--fallback-model`	Auto-fallback on overload (print mode only)	`claude -p --fallback-model sonnet "query"`
`--init / --maintenance / --init-only`	Run Setup hooks for initialization or maintenance, or exit after setup	`claude -p --maintenance "query"`
`--settings`	Load settings from JSON file or string	`claude --settings ./settings.json`
`--setting-sources`	Comma-separated list of settings sources (user, project, local)	`claude --setting-sources "user,project"`
`--safe-mode`	Troubleshoot without local customization such as CLAUDE.md, skills, plugins, and hooks	`claude --safe-mode`
`--bare`	Minimal mode for scriptable calls without automatic project customization	`claude --bare -p "query"`
`--autocompact`	Set the auto-compaction window for this session only	`claude --autocompact 500k`
`--advisor`	Enable the server-side Advisor for this session	`claude --advisor opus`
`--ax-screen-reader`	Force simplified, screen-reader-friendly output	`claude --ax-screen-reader`
`--disable-slash-commands`	Disable all skills and commands for this session	`claude --disable-slash-commands`

Note

Many flags can be combined. For example: claude --model opus --verbose --add-dir src/ starts Claude with Opus model, verbose logging, and loads the src folder.

Environment Variables

Beyond CLI flags, there are numerous environment variables that control Claude Code behavior. Set them in your shell or in settings.json under the env key. Values in a settings env override shell variables. Changes apply to running sessions unless a variable is explicitly read only at startup.

Model & Performance

VariableANTHROPIC_MODEL

DescriptionSet the default model. --model and /model override it.

DefaultProvider default

VariableCLAUDE_CODE_EFFORT_LEVEL

DescriptionSet low, medium, high, xhigh, max, or auto when the selected model supports it. Overrides /effort and settings.

DefaultModel-dependent

VariableCLAUDE_CODE_MAX_OUTPUT_TOKENS

DescriptionLimit output tokens per response. A higher value reduces usable context.

DefaultModel-dependent

VariableMAX_THINKING_TOKENS

DescriptionSet a fixed thinking budget. Adaptive thinking remains active until CLAUDE_CODE_DISABLE_ADAPTIVE_THINKING is set. 0 disables thinking except for Fable, Opus 5.5, and Sonnet 5.5.

DefaultModel-dependent

VariableCLAUDE_CODE_SUBAGENT_MODEL

DescriptionModel for subagents, Agent Teams, and workflows. Accepts an alias or full ID; inherit uses the session model.

DefaultSession model


`ANTHROPIC_MODEL`	Set the default model. --model and /model override it.	Provider default
`CLAUDE_CODE_EFFORT_LEVEL`	Set low, medium, high, xhigh, max, or auto when the selected model supports it. Overrides /effort and settings.	Model-dependent
`CLAUDE_CODE_MAX_OUTPUT_TOKENS`	Limit output tokens per response. A higher value reduces usable context.	Model-dependent
`MAX_THINKING_TOKENS`	Set a fixed thinking budget. Adaptive thinking remains active until CLAUDE_CODE_DISABLE_ADAPTIVE_THINKING is set. 0 disables thinking except for Fable, Opus 5.5, and Sonnet 5.5.	Model-dependent
`CLAUDE_CODE_SUBAGENT_MODEL`	Model for subagents, Agent Teams, and workflows. Accepts an alias or full ID; inherit uses the session model.	Session model

Bash & Timeouts

VariableBASH_DEFAULT_TIMEOUT_MS

DescriptionDefault timeout for Bash commands in milliseconds

Default120,000 (2 min)

VariableBASH_MAX_TIMEOUT_MS

DescriptionUpper bound for Bash timeouts in milliseconds

Default600,000 (10 min)

VariableBASH_MAX_OUTPUT_LENGTH

DescriptionMaximum Bash output in characters. The allowed range ends at 150,000.

Default30,000

VariableMAX_MCP_OUTPUT_TOKENS

DescriptionToken limit for MCP tool output

Default25,000


`BASH_DEFAULT_TIMEOUT_MS`	Default timeout for Bash commands in milliseconds	120,000 (2 min)
`BASH_MAX_TIMEOUT_MS`	Upper bound for Bash timeouts in milliseconds	600,000 (10 min)
`BASH_MAX_OUTPUT_LENGTH`	Maximum Bash output in characters. The allowed range ends at 150,000.	30,000
`MAX_MCP_OUTPUT_TOKENS`	Token limit for MCP tool output	25,000

Feature Toggles

VariableCLAUDE_CODE_DISABLE_AUTO_MEMORY

DescriptionDisable automatic memory creation

Value1 = off

VariableCLAUDE_CODE_DISABLE_BACKGROUND_TASKS

DescriptionDisable background tasks

Value1 = off

VariableCLAUDE_CODE_DISABLE_CRON

DescriptionStop scheduled cron jobs in session

Value1 = off

VariableCLAUDE_CODE_DISABLE_1M_CONTEXT

DescriptionDisable the 1M token context window

Value1 = off

VariableCLAUDE_CODE_DISABLE_GIT_INSTRUCTIONS

DescriptionRemove git instructions from system prompt

Value1 = off

VariableCLAUDE_CODE_ENABLE_PROMPT_SUGGESTION

DescriptionControl prompt suggestions (grayed-out suggestion text)

Valuefalse = off

VariableCLAUDE_CODE_TASK_LIST_ID

DescriptionNamed task list for cross-session tasks

Valuee.g., my-project

VariableCLAUDE_CODE_DISABLE_BUNDLED_SKILLS

DescriptionDisable bundled skills for the session

Value1 = off

VariableCLAUDE_CODE_MAX_CONCURRENT_SUBAGENTS

DescriptionMaximum number of concurrent subagents

ValueDefault: 20

VariableCLAUDE_CODE_SKIP_PROMPT_HISTORY

DescriptionDisable input and transcript history

Value1 = off

VariableCLAUDE_CODE_SUBPROCESS_ENV_SCRUB

DescriptionRemove provider secrets from spawned subprocess environments

Value1 = on

VariableENABLE_TOOL_SEARCH

DescriptionControl on-demand MCP tool loading. Accepts true, false, auto, or auto:N.

ValueProvider-dependent

VariableENABLE_CLAUDEAI_MCP_SERVERS

DescriptionControl claude.ai MCP servers

Valuefalse = off

VariableDISABLE_AUTOUPDATER

DescriptionDisable automatic updates. Manual updates remain available.

Value1 = off

VariableDISABLE_UPDATES

DescriptionBlock manual updates too.

Value1 = off

VariableDISABLE_TELEMETRY

DescriptionOpt out of telemetry and feature-flag requests. This also removes Remote Control, import, schedule, and Advisor.

ValueAny non-empty value

VariableDO_NOT_TRACK

DescriptionAlternative telemetry opt-out

Value1 = off


`CLAUDE_CODE_DISABLE_AUTO_MEMORY`	Disable automatic memory creation	1 = off
`CLAUDE_CODE_DISABLE_BACKGROUND_TASKS`	Disable background tasks	1 = off
`CLAUDE_CODE_DISABLE_CRON`	Stop scheduled cron jobs in session	1 = off
`CLAUDE_CODE_DISABLE_1M_CONTEXT`	Disable the 1M token context window	1 = off
`CLAUDE_CODE_DISABLE_GIT_INSTRUCTIONS`	Remove git instructions from system prompt	1 = off
`CLAUDE_CODE_ENABLE_PROMPT_SUGGESTION`	Control prompt suggestions (grayed-out suggestion text)	false = off
`CLAUDE_CODE_TASK_LIST_ID`	Named task list for cross-session tasks	e.g., my-project
`CLAUDE_CODE_DISABLE_BUNDLED_SKILLS`	Disable bundled skills for the session	1 = off
`CLAUDE_CODE_MAX_CONCURRENT_SUBAGENTS`	Maximum number of concurrent subagents	Default: 20
`CLAUDE_CODE_SKIP_PROMPT_HISTORY`	Disable input and transcript history	1 = off
`CLAUDE_CODE_SUBPROCESS_ENV_SCRUB`	Remove provider secrets from spawned subprocess environments	1 = on
`ENABLE_TOOL_SEARCH`	Control on-demand MCP tool loading. Accepts true, false, auto, or auto:N.	Provider-dependent
`ENABLE_CLAUDEAI_MCP_SERVERS`	Control claude.ai MCP servers	false = off
`DISABLE_AUTOUPDATER`	Disable automatic updates. Manual updates remain available.	1 = off
`DISABLE_UPDATES`	Block manual updates too.	1 = off
`DISABLE_TELEMETRY`	Opt out of telemetry and feature-flag requests. This also removes Remote Control, import, schedule, and Advisor.	Any non-empty value
`DO_NOT_TRACK`	Alternative telemetry opt-out	1 = off

Authentication & Provider

VariableANTHROPIC_API_KEY

DescriptionAPI key for Anthropic

ContextRequired without subscription

VariableANTHROPIC_BASE_URL

DescriptionAlternative base URL for an Anthropic-compatible API

ContextProxy or own gateway

VariableCLAUDE_CODE_USE_BEDROCK

DescriptionUse AWS Bedrock as provider

Context1 = enabled

VariableCLAUDE_CODE_USE_VERTEX

DescriptionUse Google Vertex AI as provider

Context1 = enabled

VariableCLAUDE_CODE_USE_FOUNDRY

DescriptionUse Microsoft Foundry as provider

Context1 = enabled

VariableHTTP_PROXY / HTTPS_PROXY

DescriptionProxy configuration

ContextFor corporate networks


`ANTHROPIC_API_KEY`	API key for Anthropic	Required without subscription
`ANTHROPIC_BASE_URL`	Alternative base URL for an Anthropic-compatible API	Proxy or own gateway
`CLAUDE_CODE_USE_BEDROCK`	Use AWS Bedrock as provider	1 = enabled
`CLAUDE_CODE_USE_VERTEX`	Use Google Vertex AI as provider	1 = enabled
`CLAUDE_CODE_USE_FOUNDRY`	Use Microsoft Foundry as provider	1 = enabled
`HTTP_PROXY / HTTPS_PROXY`	Proxy configuration	For corporate networks

Tip

You can set environment variables permanently in your settings.json instead of exporting them in the shell every time, e.g., {"env": {"CLAUDE_CODE_EFFORT_LEVEL": "medium"}}

MCP (Model Context Protocol) Integration

MCP is one of Claude Code's most important extensions. With it, you can directly integrate external tools and services. The community is already delivering hundreds of MCP server implementations and constantly expanding the ecosystem.

# Add MCP server
claude mcp add --transport http github https://api.githubcopilot.com/mcp/

# Start with debug
claude --debug='mcp'

# Expose Claude Code itself as an MCP server
claude mcp serve

Tool Search can load MCP tools on demand and save context. Control it through ENABLE_TOOL_SEARCH, for example with auto, true, false, or auto:N. The best setting depends on the provider, MCP servers, and tool count.

Another feature is MCP Elicitation: MCP servers can display interactive forms or browser URLs during a task to request structured input. MCP servers can also provide their own commands in the format /mcp__<server>__<prompt>.

Skills

One of the coolest features of Claude Code? You can create your own skills that perfectly fit your workflow. Skills are the evolution of former custom commands and follow the open Agent Skills standard.

Project skills are stored in .claude/skills/ and are available to everyone in the project. Personal skills go in ~/.claude/skills/ and only apply to you. Enterprise skills are distributed organization-wide through managed settings.

---
name: test-all
description: Runs all tests and linters
argument-hint: [test-suite] [flags]
allowed-tools: Bash(npm test:*), Bash(npm run:*)
model: sonnet
disable-model-invocation: false
context: fork
---

Run the following tests:
1. npm test $ARGUMENTS
2. npm run lint
3. npm run typecheck

Show me a summary of all results.

After creating it, simply type /test-all and Claude runs your complete test suite. I have about 10 such skills for recurring tasks.

Tip

Available frontmatter fields: name, description, argument-hint, allowed-tools, model, disable-model-invocation, user-invocable, context (fork for subagent), agent (Explore, Plan, etc.) and hooks. Variables: $ARGUMENTS, $0/$1, ${CLAUDE_SKILL_DIR}, ${CLAUDE_SESSION_ID}.

Before writing a skill from scratch, check our directory of 400+ ready-made Claude Code Skills. Many recurring tasks, like test runs, reviews, or deployments, already exist there as ready-to-use SKILL.md downloads.

Troubleshooting: When Claude Code Doesn't Do What It Should

Even after hundreds of sessions, I sometimes run into problems. Here are the most common errors and their solutions.

ProblemClaude doesn't understand project

CauseWrong start folder

SolutionStart in project root, create CLAUDE.md

ProblemToken limit reached

CauseToo much context

SolutionUse /clear, /compact for smart compression

ProblemCommands don't work

CauseOutdated version

SolutionRun /doctor, then run claude update only if it recommends an update

ProblemMCP server won't connect

CauseMissing configuration

SolutionRun claude --debug=mcp for diagnosis

ProblemShift+Enter doesn't work

CauseTerminal setup missing

SolutionUse /terminal-setup in supported terminals. Warp, iTerm2, and Windows Terminal support Shift+Enter natively.


Claude doesn't understand project	Wrong start folder	Start in project root, create CLAUDE.md
Token limit reached	Too much context	Use /clear, /compact for smart compression
Commands don't work	Outdated version	Run /doctor, then run claude update only if it recommends an update
MCP server won't connect	Missing configuration	Run claude --debug=mcp for diagnosis
Shift+Enter doesn't work	Terminal setup missing	Use /terminal-setup in supported terminals. Warp, iTerm2, and Windows Terminal support Shift+Enter natively.

You can use the following debug commands for problem analysis:

# Check system status
/status

# Check installation
/doctor

# Visualize context usage
/context

# Show token costs
/cost

# Verbose mode for details
claude --verbose

# Activate debug mode
claude --debug='api,mcp'

# Session troubleshooting
/debug "my problem"

# Open configuration
/config

Warning

/doctor checks installation and configuration. Run claude update only when the diagnosis recommends it. DISABLE_UPDATES also blocks manual updates.

Frequently Asked Questions About Claude Code Commands

Use the @ symbol followed by the file path to reference files. Claude Code then offers autocomplete for paths. You can also use /add-dir to add additional working directories that are then available in the session.

Start with 5-7 basic commands: /clear, /compact, /model, /help, /status, and /exit. These let you manage sessions and context and open help. Expand your repertoire once you have mastered the basics.

Yes! Create a SKILL.md file in .claude/skills/my-skill/ for project skills or ~/.claude/skills/my-skill/ for personal skills. Use YAML frontmatter for allowed-tools, description, argument-hint, and model. This is perfect for recurring tasks in your workflow.

Use /compact to intelligently compress the context. You can optionally pass focus instructions, e.g., /compact 'Focus on the API implementation'. Alternatively, use /clear to delete the complete context for a fresh start.

Availability depends on the model, platform, plan, and provider. That is especially true for effort levels, thinking, remote features, and individual integrations. Check /model, /effort, and /help in your own session instead of relying on a universal model list.

Use 'claude mcp add --transport http' followed by a name and server URL. The official GitHub endpoint is https://api.githubcopilot.com/mcp/. Manage the configuration via /mcp. Configure Tool Search with ENABLE_TOOL_SEARCH.

Use Claude in headless mode with --output-format json or stream-json. Combine this with -p for one-time execution without interactive mode and --max-turns to limit the number of agentic turns. Example: claude -p 'Fix ESLint errors' --output-format json --max-turns 5

Claude Code Commands (100+): The Reference

Installing and Starting Claude Code

Slash Commands (Built-in)

Bundled Skills and Workflows

Bundled Workflow

Retired Commands

Keyboard Shortcuts

General Controls

Text Editing

Multi-line Input

Quick Commands

Theme and Voice Input

Transcript Viewer

CLI Commands

CLI Flags

Basic Flags

Model & System Prompt

Agents & Subagents

Permissions & Security

MCP Server & Integration

Output & Format

Directories & Session

Remote & More

Environment Variables

Model & Performance

Bash & Timeouts

Feature Toggles

Authentication & Provider

MCP (Model Context Protocol) Integration

Skills

Troubleshooting: When Claude Code Doesn't Do What It Should

Frequently Asked Questions About Claude Code Commands

Finn Hillebrandt

Similar Articles

Codex Commands: The Complete Reference

Claude Code and GDPR: Privacy Setup for Client Projects

OpenAI Codex and GDPR: Privacy Setup for Client Work

Cursor and GDPR: Privacy Settings for Client Projects

Claude Code Alternatives 2026: 12 AI Coding Tools Compared

Is Claude Code Free? Credits, Limits & Alternatives (2026)

Installing and Starting Claude Code

Slash Commands (Built-in)

Bundled Skills and Workflows

Bundled Workflow

Retired Commands

Keyboard Shortcuts

General Controls

Text Editing

Multi-line Input

Quick Commands

Theme and Voice Input

Transcript Viewer

CLI Commands

CLI Flags

Basic Flags

Model & System Prompt

Agents & Subagents

Permissions & Security

MCP Server & Integration

Output & Format

Directories & Session

Remote & More

Environment Variables

Model & Performance

Bash & Timeouts

Feature Toggles

Authentication & Provider

MCP (Model Context Protocol) Integration

Skills

Troubleshooting: When Claude Code Doesn't Do What It Should

Frequently Asked Questions About Claude Code Commands

How do I add files to context in Claude Code?

How many Claude Code commands should I learn as a beginner?

Can I create my own Claude Code commands?

What do I do when Claude Code reaches the token limit?

Do all Claude Code commands work with every model?

How do I install MCP servers for Claude Code?

How can I integrate Claude Code into my CI/CD pipeline?

Finn Hillebrandt

Similar Articles

Codex Commands: The Complete Reference

Claude Code and GDPR: Privacy Setup for Client Projects

OpenAI Codex and GDPR: Privacy Setup for Client Work

Cursor and GDPR: Privacy Settings for Client Projects

Claude Code Alternatives 2026: 12 AI Coding Tools Compared

Is Claude Code Free? Credits, Limits & Alternatives (2026)