Skip to content

MeriaApp/openclaw-setup

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

OpenClaw Safe Setup

Most OpenClaw setups are insecure out of the box. openclaw onboard doesn't create a dedicated user, doesn't lock file permissions, doesn't set DM allowlists, and doesn't run a security audit. In early 2026, researchers found thousands of exposed OpenClaw instances and over 1,000 malicious skills on ClawHub. This repo exists because the defaults aren't safe enough.

Drop this folder into Claude Code and say "set up OpenClaw using this guide." Claude walks you through every step, checks online for the latest versions, and configures everything. Takes about 10 minutes.


What This Does (That openclaw onboard Doesn't)

openclaw onboard This repo
Dedicated non-admin user No Yes
Loopback-only networking No Yes
DM allowlist No Yes
File permission lockdown No Yes (700/600)
Config immutability No Yes (chflags uchg)
Security scanner skill No Yes (installed first)
15-point verification No Yes
Auto-checks for latest versions No Yes (via Claude Code)

Plus:

  • Token-based gateway authentication
  • launchd daemon with auto-start and crash recovery
  • 4 verified skills pre-installed
  • Bonjour/mDNS disabled, secrets redacted from logs

What You Need

  • Mac Mini M4 (or any always-on Mac with Apple Silicon)
  • 16GB RAM
  • macOS 15+
  • Admin access
  • One LLM API key (Anthropic, OpenAI, or Google)
  • A Telegram account

Quick Start (with Claude Code)

Open Claude Code in this directory and say:

"Set up OpenClaw on this machine using the CLAUDE.md guide"

Claude reads CLAUDE.md, runs each script in order, asks for your API keys, and verifies the setup. Takes about 10 minutes.

Manual Setup

# cd into the setup folder first
cd openclaw-setup

# 1. Install prerequisites (Node.js, OpenClaw)
bash scripts/prerequisites.sh

# 2. Create dedicated user (requires sudo)
sudo bash scripts/create-user.sh

# 3. Security hardening + base config
sudo -u openclaw -i bash $(pwd)/scripts/harden.sh

# 4. Configure API keys (interactive prompts)
sudo -u openclaw -i bash $(pwd)/scripts/setup-keys.sh

# 5. Set up Telegram
sudo -u openclaw -i bash $(pwd)/scripts/setup-telegram.sh

# 6. Install daemon (auto-start on boot)
sudo -u openclaw -i bash $(pwd)/scripts/install-daemon.sh

# 7. Install safe skills
sudo -u openclaw -i bash $(pwd)/scripts/install-skills.sh

# 8. Verify everything
sudo -u openclaw -i bash $(pwd)/scripts/verify.sh

# 9. Optional: daily briefing
sudo -u openclaw -i bash $(pwd)/scripts/setup-briefing.sh

Security Posture

Layer Setting
User Dedicated non-admin openclaw account
Network Loopback only (127.0.0.1)
Auth Token-based gateway
DMs Allowlist (only your Telegram ID)
Sessions Per-channel-peer isolation
Filesystem Workspace-only access
Exec Deny by default
Config Immutable (chflags uchg)
Discovery mDNS/Bonjour disabled
Logging API keys auto-redacted
Daemon Auto-start, auto-restart, crash recovery

File Structure

openclaw-setup/
├── CLAUDE.md                    # Instructions for Claude Code
├── README.md                    # This file
├── config/
│   ├── openclaw.json            # Secure base config (copied during setup)
│   └── ai.openclaw.gateway.plist  # Daemon template (reference only)
└── scripts/
    ├── prerequisites.sh         # Install Node.js + OpenClaw
    ├── create-user.sh           # Create dedicated macOS user
    ├── harden.sh                # Security hardening + config
    ├── setup-keys.sh            # API key configuration
    ├── setup-telegram.sh        # Telegram bot + allowlist
    ├── install-daemon.sh        # launchd daemon setup
    ├── install-skills.sh        # Install verified skills
    ├── setup-briefing.sh        # Daily morning briefing (optional)
    └── verify.sh                # Post-setup verification

Adding More Skills

# Inspect before installing
npx clawhub inspect <skill-name>

# Install
npx clawhub install <skill-name>

# Useful additions:
npx clawhub install shopify          # E-commerce monitoring
npx clawhub install supabase         # Database queries
npx clawhub install n8n              # Workflow automation
npx clawhub install alpaca-trading   # Stock trading (paper mode!)
npx clawhub install x-twitter        # Social monitoring
npx clawhub install github-mcp       # GitHub integration

Always run openclaw security audit --deep after installing new skills.

Maintenance

Task Frequency Command
Update OpenClaw Monthly npm update -g openclaw
Update skills Monthly npx clawhub update --all
Security audit Weekly openclaw security audit --deep
Check logs As needed tail -f ~/.openclaw/logs/openclaw.log
Check CVEs Monthly Search "OpenClaw CVE 2026"

Why Mac Mini

Factor Why
Always-on Runs as a daemon — boots, runs, restarts automatically
Power ~5W idle, ~$1/month electricity
Silent Fanless at agent workloads
Isolation Separate machine, separate account, separate keychain
Price M4 16GB is $500

Known Issues

  • CVE-2026-25253 (patched in 2026.1.29): One-click RCE via WebSocket hijacking. The install script checks for this.
  • ClawHub malware (Feb 2026): 1,184 malicious skills were purged. Only install verified skills. The security scanner checks for this.
  • macOS permissions: After granting Full Disk Access / Accessibility, you must restart Terminal. The setup reminds you.

License

MIT

About

Secure, scripted OpenClaw setup for Mac Mini. Drop into Claude Code and say 'set this up.'

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages