CLI
CLI Reference
Running grok with no arguments starts the interactive TUI. This page lists the subcommands and the flags you are most likely to use; run grok --help or grok <subcommand> --help for the complete set.
Subcommands
| Command | What it does |
|---|---|
grok login | Sign in. --device-auth uses device-code authentication for headless or remote environments |
grok logout | Sign out and clear cached credentials |
grok inspect [--json] | Show the configuration Grok discovers for this directory: rules, skills, plugins, hooks, and MCP servers |
grok models | List available models |
grok mcp <list|add|remove|doctor> | Manage MCP servers — see MCP Servers |
grok plugin <list|install|uninstall|update|enable|disable|details|validate> | Manage plugins |
grok plugin marketplace <list|add|remove|update> | Manage marketplace sources |
grok sessions <list|search|delete> | List, search, or delete sessions — see Sessions |
grok export <session-id> [output] | Export a session transcript as Markdown |
grok import [targets...] | Import sessions from Claude Code |
grok memory clear [--workspace|--global|--all] | Clear cross-session memory files |
grok worktree <list|show|rm|gc> | Manage git worktrees created for sessions — see Worktrees |
grok dashboard | Open the Agent Dashboard |
grok agent stdio | Run as an ACP agent over stdin/stdout — see Headless & Scripting |
grok wrap <command...> | Run a command in a local PTY that forwards OSC 52 clipboard writes — see Terminal Support |
grok update | Check for updates or install a specific version (--check, --version <V>, --alpha, --stable) |
grok version | Print version information |
grok completions <shell> | Generate shell completion scripts |
grok setup | Fetch and install managed configuration |
Common flags
Flags for headless runs (-p, --output-format, and related) are covered in Headless & Scripting.
| Flag | What it does |
|---|---|
--cwd <PATH> | Working directory |
-r, --resume [<ID>] | Resume a session by ID, or the most recent if omitted |
-c, --continue | Continue the most recent session for the current directory |
-s, --session-id <UUID> | Use a specific UUID for a new session |
--fork-session | When resuming, fork into a new session ID |
-w, --worktree [<NAME>] | Start the session in a new git worktree |
--ref <REF> | Branch, tag, or commit to base the worktree on |
-m, --model <MODEL> | Model ID to use |
--effort <LEVEL> | Reasoning effort |
--always-approve | Auto-approve all tool executions (alias --yolo) |
--allow <RULE>, --deny <RULE> | Permission rules — see Enterprise Deployments |
--sandbox <PROFILE> | Sandbox profile — see Enterprise Deployments |
--rules <TEXT> | Extra rules appended to the system prompt |
--system-prompt-override <TEXT> | Replace the system prompt entirely |
--tools <LIST>, --disallowed-tools <LIST> | Allow or remove built-in tools |
--max-turns <N> | Maximum number of agent turns |
--no-plan, --no-subagents, --no-memory, --disable-web-search | Disable a feature for this session |
--experimental-memory | Enable cross-session memory |
--oauth | Use OAuth when the welcome screen starts authentication |
Claude Code flag names are accepted as aliases where they overlap: --allowedTools, --disallowedTools, --append-system-prompt, --system-prompt, and --dangerously-skip-permissions.
Last updated: July 4, 2026