Reference
Command reference
Every Claude Code command in one place, each with the date it was last checked against the documentation. Removed ones stay, with the version they disappeared in, because you still meet them in other people's guides and in your own scripts.
182 items · 2 removed · The whole reference was checked on August 22, 2026.
No command matches the filter.
Session and conversation
/add-dir<path>
Adds another working directory so Claude can read and edit files outside the one you started in.
/background[prompt]
Detaches the session so it keeps working as a background agent.
/branch[name]
Creates a branch of the conversation so you can try a different direction without losing this one.
/cd<path>
Moves the session to a different working directory.
/clear[name]
Starts a new conversation with empty context.
Exits the CLI.
/fork[prompt]
Copies the current conversation into a new background session.
Continues the local session from another device.
Returns to an earlier conversation.
Rolls both the code and the conversation back to an earlier checkpoint.
Pulls a web session down into the terminal.
claude[query]
Starts an interactive session in the current directory, with an optional first prompt.
claude project purge[path]
Deletes all local Claude Code state for a project: transcripts, task lists, debug logs, file-edit history, prompt history, and the project's entry in ~/.claude.json.
Starts a Remote Control server so you can drive Claude Code from claude.ai or the Claude app, with no local interactive session.
Starts the session as a background agent and returns to the shell right away, printing the session ID.
--cloud<task|session>
Creates a new web session on claude.ai from a task description, or targets an existing one when you pass its ID or URL.
Loads the most recent conversation in the current directory, without asking which one.
--environment<environment-id>
Creates a cloud session that runs on a self-hosted environment, named by an ID that starts with ccpool_.
--exec<command>
Runs a shell command as a PTY-backed background job instead of starting a Claude session.
When resuming, gives the conversation a new session ID instead of writing back into the original.
--from-pr<pr>
Opens the session picker filtered to sessions linked to one pull request.
--name<name>
Gives the session a display name, shown in /resume and in the terminal title.
Keeps the session off disk, so nothing is saved and nothing can be resumed. Print mode only.
--ref<branch>
With --environment, bases the new session's checkout on a named ref instead of your local HEAD.
Deprecated alias for --cloud, including its existing-session form.
--remote-control[name]
Starts an interactive session with Remote Control on, so the same session can also be driven from claude.ai or the Claude app.
--remote-control-session-name-prefix<prefix>
Sets the prefix for automatically generated Remote Control session names.
--resume[session]
Resumes a session by ID or name, or opens a picker when you give neither.
--session-id<uuid>
Uses a session ID you choose for the conversation, which must be a valid UUID.
Resumes a web session in your local terminal.
Creates a tmux session for the worktree.
--worktree[name|#pr]
Starts the session in an isolated git worktree at <repo>/.claude/worktrees/<name>.
Model and configuration
/advisor[model|off]
Turns the server-side advisor tool on for this session, or off.
/autocompact[auto|<tokens>]
Sets the window at which the conversation compacts itself.
/color[color|default]
Sets the colour of the prompt bar.
/config[key=value ...]
Opens the settings interface, or sets a key straight from the command line.
/effort[level|auto|status]
Sets the effort level, from low to xhigh, plus max, ultracode and auto.
/fast[on|off]
Toggles fast mode.
Opens the keyboard shortcuts file for editing.
/model[model]
Switches the model for this session.
Configures the colour theme.
--advisor<model>
Turns on the server-side advisor tool for this session, with a model alias or a full model ID.
--autocompact<auto|tokens>
Sets the auto-compact window for this session without touching your saved settings.
--betas<headers>
Adds beta headers to the API requests this session makes.
--effort<level>
Sets the effort level for the session: low, medium, high, xhigh, max, or ultracode.
--fallback-model<models>
Names the model, or the comma-separated chain of models, to fall back to when the primary one is overloaded or gone.
--model<model>
Sets the model for this session, by alias (opus, sonnet, haiku, fable) or by full model ID.
Context
/btw[question]
Asks a side question without adding it to the conversation.
/compact[instructions]
Summarizes the conversation to free up context.
/context[all]
Shows what is currently filling the context window.
/copy[N]
Copies the last assistant response, or the Nth from the end.
/export[filename]
Exports the conversation as plain text.
Creates a CLAUDE.md guide for the project.
Edits CLAUDE.md files and turns automatic memory on or off.
--add-dir<paths...>
Adds working directories so Claude can read and edit files outside the one you started in.
--append-subagent-system-prompt<text>
Appends text to every subagent's system prompt, nested subagents included.
--append-system-prompt<text>
Appends your text to the end of the default system prompt.
--append-system-prompt-file<path>
Appends the contents of a file to the default system prompt.
--exclude-dynamic-system-prompt-sections
Moves per-machine sections out of the system prompt and into the first user message.
--system-prompt<text>
Replaces the entire default system prompt with text you give.
--system-prompt-file<path>
Replaces the default system prompt with the contents of a file.
Code review
/code-review[low|medium|high|xhigh|max|ultra] [--fix] [--comment] [pr#|branch|path]
Reviews the current diff, a pull request, a branch or a path.
Opens an interactive viewer for uncommitted changes.
/pr-comments[PR]Removed in 2.1.91
Pulled the comments from a pull request into the session.
Checks the pending changes for security problems.
Reviews the changed code for reuse and simplification, then applies the fixes.
claude ultrareview[target]
Runs ultrareview without an interactive session, prints the findings to standard output, and exits 1 when the review fails.
Work and tasks
/autofix-pr[prompt]
Starts a session that watches a pull request and pushes fixes.
/batch<instruction>
Decomposes a large change and runs the parts in parallel.
/deep-research<question>
Fans out web searches and synthesizes a report.
/goal[condition|clear]
Sets a condition for Claude to work toward.
/loop[interval] [prompt]
Runs a prompt or a command repeatedly on a schedule.
/plan[description]
Enters plan mode for a task.
Lists the background work of the current session.
Verifies work that has been done.
Emits subagent text and thinking in the output stream, so each subagent's transcript can be reconstructed.
Includes hook lifecycle events in the output stream.
Includes partial streaming events in the output.
Runs Setup hooks with the init matcher before the session starts. Print mode only.
Runs Setup and SessionStart hooks and then exits, without starting a conversation.
--input-format<text|stream-json>
Chooses the input format in print mode: text or stream-json.
--json-schema<schema>
Requires the print-mode answer to be JSON that validates against a schema you supply.
Runs Setup hooks with the maintenance matcher before the session starts. Print mode only.
--max-budget-usd<amount>
Caps how many dollars a print-mode run may spend on API calls before it stops.
--max-turns<n>
Limits how many agentic turns a print-mode run may take, and errors out when the limit is reached.
--output-format<text|json|stream-json>
Chooses the output format in print mode: text, json, or stream-json.
Prints the answer and exits, without the interactive session.
Emits a predicted next user prompt after each turn that produces one.
Re-emits user messages that came in on standard input back on standard output, as acknowledgment.
Diagnosis
/bug[report]
Reports a bug or shares the conversation.
/debug[description]
Turns on debug logging and helps track a problem down.
Runs a setup checkup and fixes what it can.
/feedback[report]
Sends product feedback.
Writes a heap snapshot for diagnosing memory use.
Builds an HTML report about recent sessions.
Writes a one-line summary of the current session.
Opens the changelog in the version picker.
Shows the current status.
Shows token usage for the session.
Prints read-only installation and settings diagnostics without starting a session.
Renders flat, screen-reader friendly output, without decorative borders or animations.
Skips auto-discovery of hooks, skills, commands, subagents, plugins, MCP servers, auto memory, and CLAUDE.md so a scripted call starts faster.
--debug[categories]
Turns on debug output, optionally narrowed to categories such as mcp or startup.
--debug-file<path>
Writes debug logs to a file path you choose, and turns on debug mode along the way.
Starts with every customization disabled, to find out whether your own configuration is the problem.
Shows the full turn-by-turn output instead of the condensed view.
Prints the version number and exits.
Agents
Manages subagent configurations.
Lists subagents, team members and sessions.
Hands a side task to a subagent.
Opens agent view, where you watch and dispatch parallel background sessions.
claude attach<id>
Attaches this terminal to a background session so you can follow it and steer it.
Prints the background-session supervisor's state, version, socket directory, and worker count.
claude daemon stop--any [--keep-workers]
Stops the background-session supervisor and the sessions it hosts.
claude logs<id>
Prints recent output from a background session without attaching to it.
claude respawn<id>
Restarts a background session, running or stopped, with its conversation intact.
claude rm<id>
Removes a background session from the list.
claude stop<id>
Stops a background session.
--agent<name>
Runs the session as a named agent, overriding the agent setting.
--agents<json>
Defines subagents inline, as JSON, for this session only.
--teammate-mode<mode>
Chooses how agent team teammates are displayed: in-process, auto, tmux, or iterm2.
Permissions and security
Scans your transcripts and builds an allowlist of the read-only calls you keep approving.
Manages the allow, ask and deny rules for tool permissions.
Shows and updates privacy settings.
Prints the built-in auto mode classifier rules as JSON.
Restores the default auto mode configuration by removing the autoMode section from your user settings.
--allow-dangerously-skip-permissions
Adds bypassPermissions to the Shift+Tab cycle without starting the session in it.
--allowedTools<rules...>
Lists the tools and call patterns that run without asking for permission.
--dangerously-skip-permissions
Skips permission prompts, which is the same as starting in bypassPermissions.
--disallowedTools<rules...>
Lists deny rules, which either remove a tool outright or refuse only the calls that match.
--enable-auto-modeRemoved in 2.1.111
Turned on auto mode for the session, before auto mode became part of the standard permission cycle.
--permission-mode<mode>
Starts the session in a permission mode: default, acceptEdits, plan, auto, dontAsk, bypassPermissions, or manual.
--permission-prompt-tool<tool>
Names an MCP tool that answers permission prompts when nobody is at the terminal.
--tools<names>
Restricts which built-in tools exist for the session: an empty value for none, default for all, or a list of names such as Bash,Edit,Read.
Integrations and environment
Configures Claude in Chrome.
Authorizes access to the design system.
/design-sync[hint]
Converts a React design system into Claude Design.
Shows the hook configuration.
Manages IDE integrations.
/import[codex|gemini] [--dry-run] [--yes]
Imports configuration from another coding agent.
Installs the Claude GitHub App.
Installs the Claude Slack app.
/mcp[reconnect <server>|enable|disable [<server>|all]]
Manages Model Context Protocol server connections.
/plugin[subcommand]
Manages plugins.
Starts the self-hosted Claude apps gateway, which puts SSO and policy in front of Claude Code on Amazon Bedrock, Google Cloud's Agent Platform, or Microsoft Foundry.
claude import[codex|gemini]
Starts a session that runs /import to bring configuration from another coding agent into Claude Code.
Configures Model Context Protocol (MCP) servers from the terminal: add them, list them, remove them.
claude mcp login<name>
Runs a configured MCP server's OAuth flow without opening the interactive /mcp panel.
claude mcp logout<name>
Clears the stored OAuth credentials for an MCP server.
Manages plugins from the terminal: installs them, removes them, and lists what is enabled.
Registers this machine or container with a self-hosted environment and hosts Claude Code cloud sessions on your own infrastructure.
--channels<plugin:name@marketplace ...>
Names the MCP servers whose channel notifications the session should listen for. Research preview.
Turns on the Chrome integration for web automation and testing.
--dangerously-load-development-channels
Enables channels that are not on the approved allowlist, for local development.
Connects to your IDE at startup when exactly one valid IDE is available.
--mcp-config<paths|json>
Loads MCP servers from JSON files or inline strings, space-separated.
Turns the Chrome integration off for this session.
--plugin-dir<path>
Loads a plugin from a directory or a .zip archive, for this session only.
--plugin-url<url>
Fetches a plugin .zip archive from a URL, for this session only.
--setting-sources<user,project,local>
Chooses which setting sources load: user, project, local, comma-separated.
--settings<path|json>
Loads settings from a JSON file or an inline JSON string for this session.
Uses only the MCP servers from --mcp-config and ignores every other MCP configuration.
Account and apps
Continues the session in the Claude Code desktop app.
Signs in to your Anthropic account.
Signs out.
Shows a QR code for the Claude mobile app.
Shares a free week with a friend, if you are eligible.
Upgrades the plan, for users outside Enterprise.
Signs you in to your Anthropic account from the terminal.
Signs you out of your Anthropic account.
Prints authentication status as JSON, exiting 0 when you are signed in and 1 when you are not.
claude install[version]
Installs or reinstalls the native binary, at a named version if you give one.
Generates a long-lived OAuth token for CI and scripts, and prints it once.
Updates the installed Claude Code to the latest version.
Other
Lists and manages artifacts.
/claude-api[migrate|managed-agents-onboard|prompt-audit]
Loads reference material for the Claude API.
/dataviz[request]
Gives design guidance for charts and dashboards.
Toggles focus view, in fullscreen only.
Shows help and the available commands.
Teaches features through short interactive lessons.
Opens the Claude FM lo-fi radio.
Turns off all skills and commands for the session.