How do I install OpenClaw?

You can install OpenClaw manually using Docker (docker pull ghcr.io/openclaw/openclaw:latest) and configuring it via JSON config files, or use OpenClaw Launch for managed hosting that deploys in 30 seconds with no setup required.

Can I install OpenClaw on Mac?

Yes. Install Docker Desktop on your Mac, then pull the OpenClaw Docker image. Alternatively, use OpenClaw Launch to skip local installation entirely — your bot runs in the cloud.

Do I need Docker to run OpenClaw?

For self-hosted installations, yes — OpenClaw runs inside a Docker container. With OpenClaw Launch, Docker is handled for you automatically. No Docker knowledge needed.

← Home

Guide

How to Install OpenClaw

The complete guide to setting up OpenClaw — from manual Docker install to one-click deploy with OpenClaw Launch.

What is OpenClaw?

OpenClaw is an open-source AI agent framework with 15,000+ GitHub stars. It lets you run AI chatbots on Telegram, Discord, or directly in your browser. OpenClaw supports models from OpenAI (GPT-5.2), Anthropic (Claude), Google (Gemini), DeepSeek, and more — plus over 5,700 community-built skills on ClawHub.

There are two ways to install and run OpenClaw: the traditional DIY approach (Docker + CLI) or the managed approach with OpenClaw Launch.

Option A: Install OpenClaw Yourself (DIY)

The traditional installation requires a server, Docker, command-line access, and manual configuration. Here are the steps:

1. Get a server or Mac Mini

You need a machine that runs 24/7. Most people rent a VPS running Ubuntu (Hetzner, DigitalOcean, Vultr — $5–20/month), buy a Mac Mini ($500+), or use a Windows desktop with WSL2. The server needs at least 1 GB RAM and a public IP address. Ubuntu is the most common choice for VPS hosting.

2. Install Docker

OpenClaw runs inside a Docker container. On Ubuntu/Debian, run: sudo apt update && sudo apt install docker.io docker-compose-plugin. On macOS, install Docker Desktop. Make sure Docker is running before continuing.

3. Pull the OpenClaw image

Run: docker pull ghcr.io/openclaw/openclaw:latest — this downloads the latest OpenClaw release. The image is around 500 MB.

4. Create the config file

OpenClaw reads its config from ~/.openclaw/openclaw.json. You need to create this file and configure your AI model provider (API key), chat platform (Telegram token or Discord token), gateway authentication, and agent settings. The JSON schema has nested objects for models, channels, plugins, agents, and gateway settings. A single typo or missing field can prevent the bot from starting.

5. Run the container

Start OpenClaw with: docker run -d --name openclaw -p 18789:18789 -v ~/.openclaw:/home/node/.openclaw ghcr.io/openclaw/openclaw:latest. You need to bind-mount the config directory and expose the gateway port. File permissions matter — the container runs as uid 1000.

6. Set up HTTPS (SSL)

For Telegram webhooks and secure browser access, you need HTTPS. Install a reverse proxy like Caddy or Nginx, point your domain to your server, and configure SSL certificates. Caddy handles this automatically, but Nginx requires manual Let's Encrypt setup.

7. Configure firewall & monitoring

Open port 443 (HTTPS) on your firewall. Set up a process manager (PM2 or systemd) to auto-restart OpenClaw if it crashes. Optionally configure log rotation and health-check scripts.

Total time: 2–6 hours depending on experience. Debugging config errors, Docker networking issues, and SSL setup are the most common time sinks.

Option B: Skip the Install with OpenClaw Launch

OpenClaw Launch is the managed hosting platform for OpenClaw. Instead of setting up Docker, config files, and SSL yourself, you use a visual editor and click deploy.

	DIY Install	OpenClaw Launch
Server required	Yes — Mac Mini or VPS	No — fully managed
Docker knowledge	Required	Not needed
Config format	JSON file (error-prone)	Visual editor
SSL / HTTPS	Manual setup	Automatic
Time to deploy	2–6 hours	Under 30 seconds
Monitoring	DIY scripts	Built-in dashboard
Cost	$5–20/mo (VPS) + time	From $3/mo

How OpenClaw Launch Works

Configure — Pick your AI model (Claude, GPT, Gemini, etc.) and chat platform (Telegram, Discord, or browser) from a visual form. No config files to edit.
Deploy — Click one button. Your OpenClaw instance launches in under 30 seconds. Docker, networking, and SSL are handled automatically.
Connect — Your AI agent is live. Chat on Telegram, Discord, or the built-in Web UI. Monitor everything from the dashboard.

Common Installation Issues (DIY)

If you go the DIY route, here are the issues people run into most:

Config syntax errors — A missing comma or wrong nesting in openclaw.json prevents the bot from starting. The error messages aren't always clear.
Docker permissions — The container runs as user "node" (uid 1000). Bind-mounted directories need the right permissions or the config can't be read.
Gateway auth format — The gateway.auth field must be an object ({ token: "..." }), not a string. Many tutorials get this wrong.
Plugin config — Both the channel config AND the plugin entry must be enabled for Telegram/Discord to actually start. Channel config alone isn't enough.
SSL/HTTPS — Telegram webhooks require HTTPS. Without a reverse proxy and valid certificate, the bot can't receive messages.

OpenClaw Launch eliminates all of these issues. The visual editor generates valid config automatically, and SSL is included out of the box.

Which Option Should You Choose?

Choose DIY if you want full control over your server, already have Docker experience, and don't mind spending a few hours on setup and ongoing maintenance.

Choose OpenClaw Launch if you want to get your AI agent running immediately, don't want to manage servers, or are new to OpenClaw. Plans start at $3/month with zero infrastructure setup.

Want to learn more about what your agent can do once it's running? Read our OpenClaw AI Agent guide.