OpenClaw CLI Cheat Sheet: Commands, Config & Skills

OpenClaw runs AI agents through a Gateway, a terminal UI, and connected chat apps. The openclaw CLI handles setup, models, configuration, Skills, and the services that keep agents running.

This OpenClaw CLI cheat sheet puts everyday terminal commands, chat slash commands, configuration keys, and Skills operations in one reference. Find the operation you need, then copy its command or adapt a short example.

Read More: 7 Best OpenClaw Alternatives for Safe & Local AI Agents

Last updated: October 18, 2026.

Essential OpenClaw CLI Commands

Global Flags and JSON Output

Installation and Initial Setup

Install OpenClaw

macOS, Linux, or WSL2:

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

Windows PowerShell:

iwr -useb https://openclaw.ai/install.ps1 | iex

The installer launches onboarding and provisions Node when needed. If you prefer to keep OpenClaw and Node under a local prefix instead of using a system-wide Node installation, use the local-prefix installer:

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

Manual global npm installation:

npm install -g openclaw@latest --allow-scripts=openclaw

Setup Commands

Gateway, Dashboard, and TUI

For an explicit remote TUI connection, supply the target credentials and complete device pairing when requested. Local mode cannot be combined with remote connection flags.

openclaw tui --url wss://gateway.example.com --token <gateway-token>

Chat Slash Commands

Help, Status, and Context

Sessions and Active Runs

Models, Thinking, and Output

Permissions and Administrative Commands

/exec has two different scopes: host/node placement persists, while security and ask apply only to the message that contains them. Include those permission options with the task they should govern.

/exec host=<auto|sandbox|gateway|node> security=<deny|allowlist|full> ask=<off|on-miss|always> node=<id> <task>

Models and Authentication

Configuration, Files, and Environment Variables

Read and Edit Configuration

The default configuration file is ~/.openclaw/openclaw.json. It accepts JSON5, including comments and trailing commas. Use openclaw config file to find the active path when profiles or environment variables change it.

Before using CLI config set, config patch, or config unset, stop the local Gateway through its service owner and wait for it to exit. These writes require exclusive state ownership. Validate the result before starting the Gateway. For a running Gateway, use its Control UI configuration editor or an authorized chat configuration command.

JSON5 Configuration Example

Merge these fields into an existing configuration to select a local, loopback-only Gateway and a default workspace. Preserve the authentication, model, and channel settings created during onboarding. For direct file edits, validate the saved file and check whether the affected setting requires a Gateway restart.

{
  gateway: {
    mode: "local",
    bind: "loopback",
    port: 18789,
  },
  agents: {
    defaults: { workspace: "~/.openclaw/workspace" },
  },
}

Configuration Keys

Important Paths

Environment Variables and Precedence

Environment loading normally preserves existing values. Precedence is: process environment, current-directory .env, global .env, the config env block, then optional login-shell import.

Provider credentials and protected runtime controls from a workspace .env are ignored. Use the global .env or the Gateway service environment for those values.

An OpenClaw-managed systemd service has a narrow exception: trusted dotenv values can replace service values that OpenClaw recorded as managed, while operator-owned service values retain precedence.

Skills and ClawHub

Find, Install, and Check Skills

Invoke or Disable a Skill

Inside chat, /skill <name> [input] invokes a user-invocable skill by name. A Skill may also have a generated direct slash command. In Control UI, selecting a Skill from slash completion inserts a $skill-name reference into the message.

/skill <name> <task or input>

To disable a file-backed skill, stop the local managed Gateway, change its enabled setting, validate, and start the Gateway again:

openclaw gateway stop
openclaw config set skills.entries.<name>.enabled false --strict-json
openclaw config validate
openclaw gateway start

Remove a ClawHub Skill

Install the standalone ClawHub CLI if needed, then choose the same workspace root used during installation. Uninstall asks for confirmation before deleting the skill directory and its lock-file entry.

npm i -g clawhub
clawhub --workdir <agent-workspace> uninstall @owner/<slug>

Create a Local Skill

Create <workspace>/skills/project-check/SKILL.md with YAML frontmatter and instructions such as the following.

---
name: project-check
description: Review a project's test commands and summarize failures.
user-invocable: true
---
# Project Check
Read the project instructions and package scripts.
Run the documented test command when execution is permitted.
Report failed tests and the files that need attention.

Publish a Skill to ClawHub

Authenticate with ClawHub before publishing a local Skill directory containing SKILL.md. Published ClawHub Skills use the MIT-0 license, allowing reuse and redistribution without attribution.

clawhub login
clawhub skill publish <path>

Agents, Sessions, and Memory

Use agent for an individual task and agents to manage configured agents. A workspace is a working directory and context source. Enable sandboxing when tools need enforced isolation.

Workspace Instruction and Memory Files

Channels and Messaging

Channel setup connects an account. Pairing grants a sender access to supported direct-message channels.

To send a message, specify the channel and its target syntax. This command sends to the selected destination immediately.

openclaw message send --channel discord --target channel:<channel-id> --message "Deployment complete."

Advanced Command Index

These entries identify the main operation for each advanced area. Run the relevant subcommand with --help for its full argument list and installed-plugin options.

Advanced Chat Entry Points

Command Examples

First Launch to Background Service

openclaw onboard --install-daemon
openclaw gateway status
openclaw dashboard

Change a Model for One Conversation

/models
/model <provider/model> -s
/model status

Change the Saved Shared Model

openclaw models list
openclaw gateway stop
openclaw models set <provider/model>
openclaw config validate
openclaw gateway start
openclaw models status

Install a Skill and Check Its Readiness

openclaw skills search "<query>"
openclaw skills install @owner/<slug>
openclaw skills info <name>
openclaw skills check
openclaw skills list --eligible

Diagnose Before Applying Repairs

openclaw config validate
openclaw doctor --lint
openclaw gateway status
openclaw channels status --probe
openclaw logs --limit 100

Troubleshooting

Sources

For syntax and options in your installed release, run openclaw --help or openclaw <command> --help. For more detailed reference material, use the official CLI reference, chat slash commands, configuration CLI, Skills CLI, ClawHub CLI, and installation instructions.

Related Cheat Sheets

• Claude Code Commands Cheat Sheet
• OpenAI Codex Commands Cheat Sheet