The OpenClaw Cheatsheet

One-Liner CLI

# macOS / Linux — installs Node.js + everything
curl -fsSL https://openclaw.ai/install.sh | bash

# Windows (PowerShell)
iwr -useb https://openclaw.ai/install.ps1 | iex

# npm global install
npm i -g openclaw

# Onboard — interactive wizard
openclaw onboard --install-daemon

From Source Advanced

git clone https://github.com/openclaw/openclaw.git ~/.openclaw
cd ~/.openclaw && npm install && npm run build && npm run init
openclaw --version

Requirements

• Node.js 22+ (node --version)
• macOS 14+, Linux, or Windows (WSL2)
• AI API key — Anthropic (recommended), OpenAI, or OpenRouter
• Low RAM? export NODE_OPTIONS="--max-old-space-size=4096"

Post-Install

openclaw --version          # Version
openclaw doctor              # Health check
openclaw doctor --deep --yes # Deep scan + auto-fix
openclaw test model          # Test AI
openclaw dashboard            # Web UI

What Goes Where Tip

• AGENTS.md → stable rules, not temp tasks
• SOUL.md → personality & ethics, not tickets
• USER.md → personalization (tone, formatting)
• TOOLS.md → guidance only, doesn't control access
• MEMORY.md → curated facts, not transcripts

Git Backup CLI

cd ~/.openclaw/workspace
git init && git add AGENTS.md SOUL.md TOOLS.md \
  IDENTITY.md USER.md HEARTBEAT.md memory/
git commit -m "Add workspace"
git remote add origin <private-repo>
git push -u origin main

Never commit: API keys, OAuth tokens, ~/.openclaw/ config.

Config CLI CLI

openclaw config get <key>
openclaw config set <key> <val>
openclaw config unset <key>
openclaw gateway restart  # Required after changes

Model Fallbacks Config

agent: { model: {
  primary: "anthropic/claude-sonnet-4-5",
  fallbacks: ["anthropic/claude-opus-4-6", "openai/gpt-5.2"]
}}

Auto-rotates auth profiles on rate limits.

Gateway CLI

openclaw gateway status|start|stop|restart
openclaw dashboard         # Web UI
openclaw tui               # Terminal UI

Channels CLI

openclaw channels list
openclaw channels status --probe
openclaw channels login          # WhatsApp QR
openclaw channels add --channel telegram --token $TOK
openclaw channels logs --channel whatsapp

Agents CLI

openclaw agents list --bindings
openclaw agents add|delete <name>
openclaw agents set-identity --from-identity

Logs & Diagnostics CLI

openclaw logs --follow        # Tail live
openclaw logs --json          # JSON
openclaw logs --limit 200
openclaw doctor --deep --yes  # Full fix
openclaw health --json --verbose

Skills & Plugins CLI

openclaw plugins list|enable|disable
clawhub install|uninstall|list|inspect <slug>
clawhub update --all

Global Flags CLI

openclaw --dev status
openclaw --profile work gateway status
openclaw status --json
openclaw logs --no-color

Directives Slash

# Per-message modifiers
:think high    # Thinking level
:model opus    # Model override
:send off      # Suppress delivery
:verbose       # Verbose mode

Skill Commands Slash

Skills auto-register as /skill-name. Use /skill <name> [input]. Auto on for Discord/Telegram.

Telegram Config

{ channels: { telegram: {
  enabled: true,
  botToken: "123456:ABC-DEF..."
}}}

Most stable — persistent API, survives reboots.

WhatsApp Config

{ channels: { whatsapp: {
  enabled: true,
  allowFrom: ["+1555..."],
  groupPolicy: "allowlist"
}}}

QR scan required. Breaks on reboot.

Discord Config

{ channels: { discord: {
  enabled: true,
  token: "BOT_TOKEN",
  allowFrom: ["USER_ID"]
}}}

Others

• iMessage — macOS only, Shortcuts bridge
• Slack — Bot + App Token + Signing Secret
• Signal — requires signal-cli
• WebChat — built-in at localhost:18789
• Extensions: MS Teams, Matrix, Zalo, Voice Call

How It Works

• Daily logs memory/YYYY-MM-DD.md — today + yesterday loaded
• Long-term MEMORY.md — curated, DM only
• Vector search — memory_search tool, ~400 token chunks
• Providers: local GGUF → OpenAI → Gemini → Voyage
• Flush at 40k tokens → distills to daily log

CLI + Tips CLI

openclaw memory index --all
openclaw memory search "query"

• Tell agent "save this to memory" for important facts
• Keep MEMORY.md compact — decisions, not transcripts
• Memory files are .md — edit directly, git-back them

SKILL.md Format File

---
name: my-skill
description: "What it does"
metadata: { "openclaw": { "requires": {
  "bins": ["uv"], "env": ["API_KEY"]
}}}
---
# Agent instructions below...

Frontmatter Keys

Priority

• Workspace (highest) — workspace/skills/
• Managed — ~/.openclaw/skills/
• Bundled (lowest) — ships with OpenClaw

Name collisions: workspace wins. Hot-reload 250ms.

How It Works

• Default: every 30m (1h Anthropic OAuth)
• Reads HEARTBEAT.md strictly
• HEARTBEAT_OK → suppressed, no notification
• Empty file → skipped (saves tokens)
• every: "0m" → disabled

Config Config

heartbeat: {
  every: "30m", target: "last",
  model: "cheap-model",
  activeHours: { start: "08:00", end: "22:00",
    timezone: "Europe/Bucharest" }
}

Job Format Config

// ~/.openclaw/cron/jobs.json
{ "name": "Morning Brief",
  "schedule": { "kind": "cron", "expr": "0 8 * * *" },
  "sessionTarget": "isolated",
  "payload": { "kind": "agentTurn",
    "message": "Deliver my briefing." }
}

Examples Tip

• 0 7 * * * — Morning briefing
• 0 18 * * 1-5 — Weekday evening recap
• 0 10 * * 1 — Weekly competitive intel
• 0 8 1 * * — Monthly invoice reminder

Cron vs Heartbeat: Cron = fixed time. Heartbeat = periodic, notifies only when needed.

Supported

Local LLM Config

models: { providers: { "ollama": {
  baseUrl: "http://localhost:11434",
  apiKey: "ollama-local",
  api: "openai-completions"
}}}

Cost Tips Tip

• Cheap model for heartbeats
• Claude Max $200/mo flat = no per-token billing
• Gemini free tier for basics
• Ollama = $0, local, private
• Edge TTS = free voice
• Set compaction.softThresholdTokens to flush early

Hardening Security

openclaw doctor --fix
openclaw security audit --deep
chmod 700 ~/.openclaw
chmod 600 ~/.openclaw/openclaw.json
chmod 700 ~/.openclaw/credentials
netstat -an | grep 18789 | grep LISTEN
grep -r "sk-" ~/.openclaw/

DM & Group Policies Config

security: {
  dmPolicy: "paired-only",
  groupPolicy: "allowlisted",
  sandboxEnabled: true
}

• allowFrom per channel for whitelisting
• per-channel-peer scope for multi-user
• Groups: requireMention: true

Skill Vetting Security

• Red flags: curl|bash, base64, Gatekeeper bypass
• Cisco scanner: pip install cisco-ai-skill-scanner
• Check for zero-width Unicode in SKILL.md
• Never install skills with auto-update behavior

Universal Fix

openclaw doctor --deep --yes

Solves 80% of issues.

Common Issues

Update CLI

sudo npm i -g openclaw@latest
openclaw update status
openclaw update wizard

TTS Providers

Auto: messages.tts.auto: "always" · Toggle: /tts on|off

Browser Automation CLI

openclaw browser fill --fields \
  '[{"ref":"name","value":"John"}]'
openclaw sandbox list|recreate|explain

Navigate, click, screenshot, extract. Needs browser.enabled: true.