Complete reference · checked against version 2.1.269

Claude Code
Command Reference

Every terminal command, every option and every slash command inside a session — with an explanation of what it does.

Compiled by Christian's virtual server · 12 September 2026 · Update of the edition of 11 September 2026 (2.1.268)

How this list came about

Terminal commands and options (parts A and B) are read straight out of the installed version 2.1.269 — claude --help and the help of every subcommand.

The slash commands (part C) come from the command definitions inside the program itself. The extraction also picks up commands whose description is only assembled at runtime or fetched from a variable, and since this edition equally those whose name sits in a variable instead of being spelled out in the code. Both are artefacts of the code being minified at build time and say nothing about the command itself.

Commands that exist in the program but are switched off by a flag are listed separately at the end, so that nothing appears available that is not.

What has changed since the last edition

Compared with 2.1.268: one new slash command, /output-style (list output styles or switch to one) — the reference now lists 139 slash commands. No command was removed, and no existing command's description changed. 37 commands had only internal, non-visible identifiers renamed, which has no effect on anyone reading the reference.

claude --help came out byte-for-byte identical. The sub-command help has one substantive change: claude plugin eval now describes in more detail what the command does and what to watch for — it loads the plugin and runs its eval suite (prompts, graders; scaffold scripts and real MCP servers only when you opt in) on your own machine, so only plugins you trust should be evaluated. The first run in an untrusted plugin directory asks for confirmation (--trust-plugin answers that for CI).

Part A · Terminal commands and options

Call form: claude [options] [command] [input]. With no arguments it starts a session in the current folder.

Starting a session

claude
Starts a session in the current folder.
claude "your question"
Starts with a first message.
claude -p "…"
Answers once and exits — for scripts.

Options

--add-dir <directories...>
Additional folders Claude is allowed to touch.
--agent <agent>
Agent for this session; overrides the setting.
--agents <json>
Define custom agents inline as JSON.
--allow-dangerously-skip-permissions
Allows turning off all prompts later, without doing so right away.
--allowedTools, --allowed-tools <tools...>
Allow only these tools, e.g. "Bash(git *) Edit".
--append-system-prompt <prompt>
Append text to the system prompt.
--autocompact <auto|tokens>
When to compact automatically (auto, or 100k–1M).
--ax-screen-reader
Screen-reader output: flat text, no borders, no animations.
--bg, --background
Start the session in the background and return right away. Prints the id that claude attach, logs, stop and rm take; claude agents lists them. With --resume , continues that session in the background under the same ID, or starts a copy and says so when the session is already running.
--bare
Frugal mode: no hooks, plugins, auto-memory or CLAUDE.md discovery.
--betas <betas...>
Send beta identifiers to the API (API key users only).
--brief
Lets Claude send you short progress notes while it works.
--chrome
Turn on the Claude-in-Chrome integration.
--cloud [description|session_id|url]
Create a cloud session or attach to an existing one.
-c, --continue
Continue the most recent conversation in this folder.
--dangerously-skip-permissions
Turn off all prompts. Only for sealed-off environments without internet.
-d, --debug [filter]
Turn on debugging, optionally filtered (e.g. "api,hooks").
--debug-file <path>
Write the debug log to this file.
--disable-slash-commands
Turn off all skills.
--disallowedTools, --disallowed-tools <tools...>
Block these tools.
--effort <level>
Thinking effort for the session: low, medium, high, xhigh, max.
--environment <environment_id>
Start a cloud session in a self-hosted environment.
--exclude-dynamic-system-prompt-sections
Keep machine-specific details out of the system prompt (better caching).
--fallback-model <model>
Backup model for when the chosen one is overloaded (now also outside --print).
--file <specs...>
Download files at startup (file_id:target_path).
--fork-session
On resume, issue a new session ID instead of continuing the old one.
--forward-subagent-text
Pass through the text and thinking of subagents as well.
--from-pr [value]
Resume the session belonging to a pull request (number or URL).
-h, --help
Show help.
--ide
Connect to the development environment automatically at startup.
--include-hook-events
Write hook events into the output stream as well (only with --print).
--include-partial-messages
Emit parts of the answer as they arrive (only with --print).
--input-format <format>
Input format: text or stream-json (only with --print).
--json-schema <schema>
JSON schema the answer is validated against.
--max-budget-usd <amount>
Dollar ceiling for the API calls (only with --print).
--mcp-config <configs...>
Load MCP servers from JSON files or strings.
--model <model>
Model for this session.
-n, --name <name>
Display name for the session.
--no-chrome
Turn off the Chrome integration.
--no-session-persistence
Do not write the session to disk.
--output-format <format>
Output format: text, json or stream-json (only with --print).
--permission-mode <mode>
Permission mode: acceptEdits, plan, bypassPermissions, default.
--permission-prompts <target>
Who answers permission prompts with --print: "host" (the SDK host or --permission-prompt-tool) or "none" (auto-deny; the permission mode still decides everything else). Default: host.
--plugin-dir <path>
Load a plugin from a folder or archive, for this session only; a folder holding several plugins loads all of them.
--plugin-url <url>
Load a plugin archive from a URL, for this session only.
-p, --print
Print the answer and exit — for scripts and pipes.
--prompt-suggestions [value]
Turn on prompt suggestions.
--remote-control [name]
Start the session with Remote Control (phone, claude.ai).
--remote-control-session-name-prefix <prefix>
Name prefix for auto-named Remote Control sessions.
--replay-user-messages
Echo incoming user messages back for acknowledgement.
-r, --resume [value]
Resume a conversation by ID, or open the picker.
--safe-mode
Start without any customisation: no CLAUDE.md, skills, plugins, hooks.
--session-id <uuid>
Use a fixed session ID.
--setting-sources <sources>
Which setting sources to load: user, project, local.
--settings <file-or-json>
Additional settings from a file or a JSON string.
--strict-mcp-config
Use only the MCP servers given via --mcp-config.
--system-prompt <prompt>
Replace the system prompt for the session.
--system-prompt-snapshot <on|off>
Record the system prompt once per conversation and reuse it verbatim on every request and resume (recommended: on). On by default for the built-in prompt; --system-prompt or --append-system-prompt turns it off so the given text applies fresh each launch.
--teleport [session]
Resume a teleport session, optionally by session ID.
--tmux
Create a tmux session for the worktree (needs --worktree).
--tools <tools...>
Pick the built-in tools: "" for none, "default" for all, or a list.
--verbose
Override the verbose setting from the configuration.
-v, --version
Show the version number.
-w, --worktree [name]
Create a new git worktree for this session (name optional).

Part B · Subcommands

Calls of the form claude <subcommand>. Indented below each one, its own subcommands.

claude agents
Manage background agents.
claude attach
Open a background session in this terminal.
claude auth
Manage sign-in.
claude auth login
Sign in to the Anthropic account.
claude auth logout
Sign out.
claude auth status
Show sign-in status.
claude auto-mode
View or reset the settings of automatic mode.
claude auto-mode config
Print the effective configuration as JSON.
claude auto-mode critique
Get feedback on your own rules.
claude auto-mode defaults
Print the built-in default rules.
claude auto-mode reset
Reset to the default rules.
claude doctor
Check the installation. For the full examination use /doctor in a session.
claude gateway
Run a company gateway for sign-in and telemetry.
claude import
Take over the configuration of another AI tool.
claude install
Install the native build (stable, latest or a fixed version).
claude logs
Show a background session's recent terminal output.
claude mcp
Set up and manage MCP servers.
claude mcp add
Add an MCP server (stdio, http or sse).
claude mcp add-from-claude-desktop
Import servers from Claude Desktop (Mac and WSL only).
claude mcp add-json
Create a server from a JSON string.
claude mcp get
Show the details of one server.
claude mcp list
List the configured servers.
claude mcp login
Sign in to a server.
claude mcp logout
Sign out of a server.
claude mcp remove
Remove a server.
claude mcp reset-project-choices
Reset the per-project approvals.
claude mcp serve
Run Claude Code itself as an MCP server.
claude plugin
Manage plugins.
claude plugin details
Contents of a plugin and its estimated context cost.
claude plugin disable
Turn a plugin off.
claude plugin enable
Turn a plugin on.
claude plugin eval
Run test cases against a plugin and score them.
claude plugin init|new
Scaffold a new plugin.
claude plugin install|i
Install a plugin from a marketplace.
claude plugin list
List the installed plugins.
claude plugin marketplace
Manage marketplaces.
claude plugin prune|autoremove
Remove plugins that are no longer needed.
claude plugin tag
Tag plugins with attributes.
claude plugin uninstall|remove
Uninstall a plugin.
claude plugin update
Update a plugin.
claude plugin validate
Check a plugin for errors.
claude project
Manage project state.
claude project purge
Delete all stored state of a project.
claude respawn
Restart background sessions so they run the current version.
claude rm
Delete a background session, and its worktree when safe.
claude setup-token
Create a long-lived authentication token (for scripts and CI).
claude stop|kill
Stop a background session; its conversation is kept.
claude ultrareview
Start a multi-stage code review in the cloud.

Part C · Slash commands inside a session

134 available commands, in alphabetical order. In grey behind them: second names and expected arguments.

/add-dir <path>
Add another working folder.
/advisor
Let a stronger model weigh in at the important moments.
/agents
Removed. Subagents now live in .claude/agents/.
/artifact-capabilities
What published pages can do at runtime.
/artifact-components
A kit of reusable building blocks for published pages.
/artifact-design
Design rules for published pages.
/artifact-diagramming
Diagrams for published pages.
/artifact-pr-review [pr number or url]
Publishes the review of a pull request as a shareable page: bottom line, recommendation, judgment calls, visual explainer, blind spots.
/artifacts
Browse published and shared artifacts.
/auto-mode-setup [--request-id <uuid>] (--wizard posture=\u2026 scope=\u2026 depth=\u2026 --propose | --expect-sha256 <64-hex> --apply-file <path>)
Set up and tune automatic mode.
/autocompact [auto|<tokens>]
Decide at which fill level the context is compacted automatically.
/autofix-pr
Watch the current pull request and fix what breaks.
/background also /bg · [prompt]
Send the session to the background and free up the terminal.
/batch <instruction>
Plan a large change, then have it carried out in parallel.
/branch [name]
Fork the conversation at this point.
/brief
Toggle brief mode.
/btw [question]
Ask a side question without disturbing the current thread.
/bug also /share · [report]
Report a bug or share the conversation.
/chrome
Open the settings for Claude in Chrome.
/claude-api
Reference for the Claude API (models, prices, tools).
/claude-code-docs [question]
Answer questions about Claude Code itself (commands, flags, settings, hooks, skills), checked against the installed build.
/claude-in-chrome
Drive the browser: click, type, read pages.
/clear also /reset, /new · [name]
New session with an empty context; the old one is kept.
/cloud-plugins
Decide whether cloud sessions use the plugins enabled here.
/code-review also /review · […] [--fix] [--comment] [<pr#>|<branch>|<path>]
Review changes for bugs and simplifications.
/color
Colour of the prompt box for this session.
/commit [guidance]
Creates a git commit, with a fitting message written for it.
/commit-push-pr
Commit, push and open a pull request — in one go.
/compact <optional custom summarization instructions>
Free up context by summarising what happened so far.
/config also /settings · [key=value]
Open the settings.
/context [all]
Show context usage as a coloured grid.
/copy
Copy the last answer to the clipboard (/copy N for older ones).
/cowork-plugin
Creates a new Cowork plugin or adapts an existing one.
/daemon
Manage background services and routines.
/dataviz
Design charts and analytics pages properly.
/debug [issue description]
Turn on debugging for this session.
/design [sync|login|consent|revoke|import|export|status|<prompt>]
Grant or revoke access to design projects.
/design-consent
Grant access to design projects.
/design-login
Unlock design access for /design-sync.
/design-revoke
Revoke access to design projects.
/design-sync
Push a React design system to claude.ai/design.
/desktop also /app
Continue the running session in the desktop app.
/diff
Look at uncommitted changes and changes per turn.
/doctor also /checkup
Examine the installation and fix problems.
/effort
Set the thinking effort.
/exit also /quit
End the session.
/explain-usage
Have your own usage explained to you.
/export [filename]
Write the conversation to a file or the clipboard.
/extra-usage
Renamed to /usage-credits.
/fast [on|off]
Toggle fast mode (same model, higher speed).
/feedback [report]
Send feedback to Anthropic.
/fewer-permission-prompts
Spot frequent harmless calls and allow them for good.
/focus
Show only the input, the summary and the answer.
/fork <directive>
Start a background agent that inherits the whole conversation.
/goal [<condition> | clear]
Set a goal Claude checks before it stops.
/heapdump
Write a memory dump (for troubleshooting).
/help
Show help and the available commands.
/hooks
Look at the hooks configured for tool events.
/ide [open]
Manage the connection to the development environment.
/import [codex|gemini|cursor] [--dry-run]
Take over the configuration of another AI tool.
/init
Generate a first CLAUDE.md from the existing code.
/insights
Build a report about your own sessions.
/install [options]
Install the native build.
/install-github-app
Set up Claude GitHub Actions for a repository.
/install-slack-app
Install the Claude app for Slack.
/keybindings
Open the keyboard shortcuts file.
/limit-reset
Reset your session limit now and keep working; available once a week, still counts toward the weekly limit.
/list-agents also /peers
List subagents and other sessions you can reach.
/login
Sign in to the Anthropic account or switch accounts.
/logout
Sign out of the Anthropic account.
/loop also /proactive
Run a task over and over at a fixed interval.
/low-priority
Continue after hitting the session limit at lower priority; run again to stop.
/mcp [reconnect <server>|enable|disable [<server>|all]]
Manage MCP servers.
/memory
Edit CLAUDE.md files and the memory settings.
/memory-types
Reference on the kinds of memory: what belongs where, and when.
/mobile also /ios, /android
Show the QR code for the phone app.
/model [model]
Set the model for Claude Code.
/output-style [style]
List output styles or switch to one.
/passes
Give away a free week of Claude Code.
/permissions also /allowed-tools
Manage the rules for allowed and forbidden tools.
/plan [open|<description>]
Turn on plan mode or look at the session's plan.
/plugin also /plugins, /marketplace
Manage plugins.
/plugin-authoring
Helps you write and debug a plugin made of function hooks.
/plugin-types [dir]
Write two type files: the plugin API's TypeScript declarations and the inputs of the connected MCP tools, for typing a hooks module against this session.
/powerup
Get to know Claude Code's features in short lessons.
/privacy-settings
View and change the privacy settings.
/pro-trial-expired
Your options once the Pro trial has run out.
/prototype
Turns an idea into a working draft and publishes it as a page.
/radio
Turn on the lo-fi radio.
/rate-limit-options
Show the options once the usage limit is reached.
/recap
Produce a one-line summary of the session.
/release-notes
Show what is new in each version.
/reload-plugins [--force]
Apply pending plugin changes.
/reload-skills
Read in newly added or changed skills.
/remote-control also /rc
Drive the session from your phone or from claude.ai.
/remote-env
Choose the default environment for cloud agents.
/rename also /name · [name]
Rename the conversation.
/resume also /continue · [conversation id or search term]
Continue an earlier conversation.
/rewind also /checkpoint, /undo
Roll code and/or conversation back to an earlier state.
/run
Start the application and check a change in real operation.
/run-skill-generator
Turn a workflow of yours into a new skill.
/sandbox
Set up and inspect the sandbox.
/schedule also /routines
Create and manage scheduled cloud agents.
/scroll-speed
Set the scroll speed of the mouse wheel.
/session also /remote
Show the address and QR code of the cloud session.
/setup-bedrock
Set up Amazon Bedrock again.
/setup-claude also /setup-cowork
Walks you through setup: pick a role, install a matching plugin, try a skill, connect tools.
/setup-vertex
Set up Google Vertex AI again.
/simplify [<target>]
Tidy up and simplify changed code.
/skill-doctor
Show which loaded skills are costing context without being used.
/skills
List the available skills.
/status
Show the state: version, model, account, connections.
/statusline
Set up the status line.
/stickers
Order Claude Code stickers.
/stop
End this background session; log and worktree are kept.
/subtask <task>
Send off a subagent with full context; the result comes back.
/tasks also /bashes
View and manage everything running in the background.
/team-onboarding
Generate a guide for colleagues out of your own usage.
/teleport also /tp
Continue a session from claude.ai here.
/terminal-setup
Set up the terminal key bindings (line break).
/theme
Switch the colour scheme.
/tui [default|fullscreen]
Choose the interface style: normal or full screen.
/ultraplan <prompt>
Have an editable plan drafted on the web.
/ultrareview
Multi-stage bug hunt across the branch, in the cloud.
/upgrade
Move to Max for higher limits and more Opus.
/usage also /cost, /stats
Show the session's cost, usage and activity.
/usage-credits
Set up credits or ask your administrator for them.
/verify
Checks that a change really does what it should — in a real run.
/voice [hold|tap|off]
Toggle voice mode.
/web-setup
Connect Claude Code on the web with your GitHub account.
/whiteboard
Opens a free drawing surface for sketches and diagrams — on your own or, where available, live together with others.
/workflow-launch-exec
Execute a workflow triggered by the server.
/workflows
Look through running and finished workflows.
/workshop
Designs something with you step by step, as a growing document.

Appendix · Present, but switched off

These names exist in the program, but are disabled by a flag in this version and are not available inside a session.

/loops
List, create and delete loops. present in the program, but switched off
/pause-memory also /memory-pause, /toggle-memory
Pause automatic memory for this session. present in the program, but switched off
/update also /restart
Move to the newest version. present in the program, but switched off
/version
Show the version of this session. present in the program, but switched off
/wellbeing also /breaks, /break-reminder, /downtime
Set up break reminders and quiet hours. present in the program, but switched off
Claude Code 2.1.269 · 64 options · 134 available slash commands · 5 switched off · as of 12 September 2026