How to Install OpenClaw: Complete Installation Guide

1. Installation routes — my recommendations

OpenClaw requires Node.js 22+ and runs on macOS, Linux, and Windows (via WSL2). I do not treat every method as equal:

What I run: learn on a laptop with the quick installer, then move the Gateway to a small VPS once a workflow (like the morning briefing) must fire every day.

2. Pre-Installation Checklist

Before installing, ensure you have:

• Node.js 22+ - OpenClaw requires Node 22 or newer. Check with node -v; install from nodejs.org or your package manager.
• OS compatibility - macOS 12+, Ubuntu 20.04+ (or equivalent Linux), or Windows 10/11 with WSL2.
• API keys - At least one LLM provider: Anthropic (Claude) or OpenAI (GPT). Optional: Google (Gemini), DeepSeek, or Ollama for local models. See the complete Ollama & offline AI guide.
• Messaging app account - For your first channel: Telegram (easiest), Discord, WhatsApp, or another supported platform.
• VPS (if applicable) - For cloud deployment: a droplet/instance with 2+ GB RAM and a public IP (or VPN/Tailscale for secure access). See network isolation and security best practices before exposing any service.

3. Quick install (one-line) — then prove it works

On macOS or Linux, the official installer downloads OpenClaw, installs dependencies, and prepares the CLI and gateway. Run:

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

Or with npm (if Node.js 22+ is already installed):

npm install -g openclaw

You should see the installer confirm the install path and prompt you to run openclaw. After installation:

# Reload shell if `openclaw` is not found
source ~/.bashrc   # or ~/.zshrc

openclaw --version
openclaw doctor
openclaw gateway status
openclaw status

Run onboarding with openclaw onboard or openclaw setup to choose your LLM provider and add your first channel. Full walkthrough: Quick Start Guide. Control UI is typically at http://127.0.0.1:18789/ when the Gateway is local.

Field note: if openclaw: command not found appears, 90% of the time the shell has not reloaded PATH. Restart the terminal before reinstalling. Prefer openclaw gateway install for a supervised always-on service (see Gateway CLI docs).

4. Platform-Specific Guides

5. First channel milestone (Telegram)

Do not customize ten skills before you get a single reply. My preferred first channel is Telegram:

1. Create a bot with BotFather and copy the token (store as env var, not in git).
2. Add the bot in OpenClaw channel config during onboarding or via config files (config overview).
3. Message the bot /ping or a short hello; confirm the agent replies.
4. Only then add WhatsApp/Discord or heavy skills.

# Conceptual — keep secrets out of shell history when possible
export TELEGRAM_BOT_TOKEN="XXXX"
# Then complete channel binding via onboarding / config as documented for your version

When that works, you have a real base for workflows like the morning briefing or PR review assistant.

6. Security basics — do these before you expose anything

• Bind the Gateway to localhost (or put it behind Tailscale/VPN)—never leave admin UI on 0.0.0.0 casually.
• SSH keys only on VPS; disable password root login.
• API keys in environment variables / secret store—not committed JSON.
• Set provider spend caps where available.
• Follow the full security checklist and security best practices.

Field note: the fastest way to regret a VPS install is an open port plus a powerful agent skill. Harden first, automate second.

7. Post-installation configuration

After OpenClaw is installed, complete these steps:

1. Run the onboarding wizard - Select an LLM provider and add your first messaging channel. Post-installation setup walks through this in detail.
2. Choose your first channel - Telegram is easiest; see Channel Setup Hub.
3. Test the connection - Send a message; check openclaw gateway status, openclaw status, and openclaw logs --follow if anything fails.
4. Security hardening - Complete section 6 above before public exposure.
5. Budget - Rough out monthly spend with the cost playbook and cost calculator.

Post-Installation Setup →

8. System requirements

9. Next steps after installation

1. Harden — Security best practices and checklist.
2. Ship one workflow — Morning briefing or PR assistant.
3. Tune cost — Cost playbook.
4. Configure memory and agents — Config and agents.
5. Install skills carefully — ClawHub + skills audit.
6. Community — Community hub; site questions via Contact / Maintainer.

10. Common installation problems

If something goes wrong, run openclaw doctor first—it checks your environment and can suggest fixes. Then refer to:

• "openclaw command not found" - PATH issue: restart the terminal or source ~/.bashrc. See Troubleshooting: installation.
• Node.js version conflicts - Requires Node 22+. Use nvm or your package manager.
• Gateway won’t start - Check ports and bind address (prefer localhost). Gateway issues.
• Channel silent - Token, webhook, or firewall; start with Telegram diagnostics in channels troubleshooting.
• Permission errors - Write access to install/config directories.
• OOM on tiny VPS - Undersized RAM with local models; right-size or use API models first (cost & sizing).

Full Troubleshooting Guide → · Installation errors