OpenClaw Reference

CLI Commands

OpenClaw is controlled entirely through the CLI. All commands follow the pattern openclaw <command> [options].

Core Commands

Command	Description
openclaw init	Initialize a new workspace and config at ~/.openclaw/
openclaw gateway start	Start the gateway daemon (main process)
openclaw gateway stop	Stop the running gateway
openclaw gateway restart	Restart the gateway
openclaw gateway status	Check gateway status and connected channels
openclaw help	Show all available commands

Session & Agent Commands

Command	Description
openclaw session list	List active sessions
openclaw session kill <id>	Terminate a specific session
openclaw system event --text "msg"	Send a system event notification

Configuration

OpenClaw uses a YAML configuration file at ~/.openclaw/config.yaml. Here's a complete example:

# ~/.openclaw/config.yaml
agent:
  name: "MyAgent"
  personality: "helpful and concise"

models:
  default: anthropic/claude-sonnet-4
  providers:
    anthropic:
      apiKey: sk-ant-...
    openai:
      apiKey: sk-...
    google:
      apiKey: AIza...
    openrouter:
      apiKey: sk-or-...
    ollama:
      baseUrl: http://localhost:11434

channels:
  telegram:
    botToken: "123456789:ABC-DEF..."
    allowedUsers:
      - 123456789
  whatsapp:
    enabled: true
  discord:
    botToken: "MTIz..."
    guildId: "98765..."

gateway:
  port: 3333
  bind: "0.0.0.0"

workspace:
  path: ~/.openclaw/workspace

Supported Models

OpenClaw supports any model accessible via these providers:

Provider	Models	Config Key
Anthropic	Claude Opus 4, Sonnet 4, Haiku 4	anthropic/claude-opus-4
OpenAI	GPT-4o, o1, o3, GPT-4o-mini	openai/gpt-4o
Google	Gemini 2.5 Pro, 2.5 Flash	google/gemini-2.5-pro
Ollama (local)	Llama 3, Mistral, DeepSeek, Qwen	ollama/llama3
OpenRouter	100+ models from all providers	openrouter/provider/model

Skills System

Skills are modular capabilities defined in Markdown files. Each skill lives in its own directory under ~/.openclaw/skills/ or ~/.npm-global/lib/node_modules/openclaw/skills/.

Skill Structure

skills/
  my-skill/
    SKILL.md      # Skill definition (required)
    scripts/      # Helper scripts
    references/   # Reference docs

SKILL.md Format

# My Custom Skill

## Description
What this skill does and when the agent should use it.

## Instructions
Step-by-step instructions the agent follows when
this skill is triggered.

## Tools Used
- exec: for running shell commands
- web_fetch: for HTTP requests
- read/write: for file operations

Workspace Files

The agent's workspace at ~/.openclaw/workspace/ is where your agent stores knowledge, tasks, and project files.

File	Purpose
SOUL.md	Agent personality, identity, and operating instructions
USER.md	Information about the user (you)
AGENTS.md	Workspace rules and agent loading behavior
TOOLS.md	Available tools and integration notes
memory/	Daily notes and long-term memory files

Deployment Options

OpenClaw runs anywhere Node.js runs:

• Mac (local): Best for personal use and development
• Linux VPS: Best for always-on 24/7 operation — see our VPS guide
• Raspberry Pi: Ultra-low-cost always-on option