How to Deploy OpenClaw

Short on time? Our Setup Kit ($499) handles all of this for you — professionally configured, verified, and documented. Async delivery, no phone calls.

Before you start, make sure your host machine meets these requirements:

• Operating system: macOS 13 (Ventura) or later, or a modern Linux distribution (Ubuntu 22.04+, Debian 12+). Windows works via WSL2.
• Node.js: Version 20 or later. OpenClaw's CLI and gateway are Node-based. Check with node --version.
• Memory: Minimum 4 GB RAM available. The gateway itself is lightweight (~200 MB), but skills that spawn subprocesses (browser automation, media processing) need headroom.
• Network: Stable internet for LLM API calls. If you're behind a corporate firewall, ensure outbound HTTPS to your model provider (OpenAI, Anthropic, etc.) is allowed.
• API keys: At least one LLM provider API key (OpenAI, Anthropic, Google, etc.). You'll configure this during gateway init.
• Optional: A domain name if you want to expose your agent via webhooks (Discord, Slack, Telegram integrations).

1Install the OpenClaw CLI

The CLI is the primary interface for managing your OpenClaw installation. Install it globally via npm:

npm install -g openclaw
openclaw --version   # verify installation

The CLI manages your gateway daemon, skills, configuration, and cron jobs. It's the single entry point for all OpenClaw operations.

2Initialize and Configure the Gateway

The gateway is the core daemon that processes messages, routes to LLMs, and executes skills. Initialize it:

openclaw gateway init

This creates the config directory (~/.openclaw/) and a default configuration file. Next, set your model provider:

# Set your primary model
openclaw config set model anthropic/claude-sonnet-4-20250514

# Add your API key
openclaw config set anthropicKey sk-ant-...

# Optional: configure additional providers
openclaw config set openaiKey sk-...

Key configuration decisions: Which model to use as default (affects cost and capability), which channels to enable (Discord, Slack, webchat), and whether to enable the workspace for file operations.

3Install Skills

Skills are modular capabilities that extend your agent. Each skill adds specific tools — web search, file management, browser control, API integrations, and more. Browse available skills:

# List available skills
openclaw skill list

# Install specific skills
openclaw skill install web-search
openclaw skill install browser
openclaw skill install github

# Verify installed skills
openclaw skill list --installed

Start with 3–5 core skills that match your use case. You can always add more later. Each skill may have its own dependencies (API keys, system packages) documented in its README. Our directory lists 116+ available skills with descriptions and requirements.

4Connect a Channel

Channels are how you interact with your agent. The built-in webchat works immediately, but you'll likely want to connect a messaging platform:

# Discord (most common)
openclaw config set discordToken YOUR_BOT_TOKEN

# Slack
openclaw config set slackBotToken xoxb-...
openclaw config set slackAppToken xapp-...

# Telegram
openclaw config set telegramToken 123456:ABC...

Each channel has its own setup requirements (creating a bot application, configuring permissions, setting webhook URLs). The OpenClaw docs cover each channel in detail.

5Start the Gateway and Verify

Launch the gateway daemon and confirm everything is working:

# Start the gateway
openclaw gateway start

# Check status
openclaw gateway status

# Test with a simple prompt
openclaw chat "What skills do you have available?"

If the gateway starts successfully and responds to your test prompt with a list of installed skills, your deployment is working. Common issues at this stage:

• API key errors: Double-check your provider key is set correctly and has sufficient credits.
• Port conflicts: The gateway binds to a local port. If something else is using it, configure a different port.
• Skill failures: Some skills require additional system dependencies (e.g., browser skill needs Chromium). Check skill-specific docs.

6Set Up Persistence (Optional but Recommended)

For production use, you'll want the gateway to survive reboots and recover from crashes:

• macOS: Create a LaunchAgent plist so the gateway starts on boot and restarts on failure.
• Linux: Create a systemd service unit with Restart=on-failure.
• Monitoring: Set up a health check cron that hits openclaw gateway status and alerts on failure.

This is where self-deployment gets time-consuming. Getting the daemon management, log rotation, health checks, and recovery right takes iteration.

• Skipping verification: Always test each skill after installation. A skill that installs without errors can still fail at runtime due to missing environment variables or dependencies.
• Too many skills at once: Start with a focused set. Each skill adds context to the system prompt, which affects token usage and response quality. Add skills incrementally.
• No workspace backup: The ~/.openclaw/workspace directory contains your agent's memory and configuration. Back it up or version-control it.
• Ignoring model costs: Different models have vastly different per-token costs. A misconfigured cron job hitting GPT-4 every minute can generate surprising bills.
• Channel permission gaps: Discord/Slack bots need specific OAuth scopes. If your agent can't read messages or post responses, check the bot's permission configuration in the platform's developer portal.