Install OpenClaw in 5 Steps: From Zero to a Working AI Assistant

Introduction

OpenClaw is an open-source AI assistant that you can fully self-host. Unlike cloud services such as ChatGPT or Claude, you retain complete control over your data and only pay for API calls to the language model of your choice. But the installation process can seem daunting if you do not work with servers on a daily basis.

In this guide, we walk through the complete installation process step by step. Whether you choose a Europese cloud VPS, a Raspberry Pi, or an old laptop tucked under your desk — after reading this article you will have a working OpenClaw instance up and running.

Step 1: Check System Requirements

OpenClaw runs on virtually any Linux machine with at least 2 GB of RAM and 10 GB of free disk space. The software itself is lightweight; the heavy lifting is done by external AI models via API calls. You need Docker (version 24 or higher) and Docker Compose. On Ubuntu, install both with sudo apt install docker.io docker-compose-v2. Verify your version with docker --version.

For networking, you need a domain name if you want to access OpenClaw over HTTPS — which is strongly recommended if you plan to connect WhatsApp or Telegram. A free domain via DuckDNS works perfectly for home use. Make sure ports 443 and 80 are open in your firewall.

Step 2: Download and Configure OpenClaw

Clone the repository with git clone https://github.com/openclaw/openclaw.git and navigate into the folder. Copy the example configuration file with cp .env.example .env and open it in your favorite editor. The most important setting is your AI provider: fill in your API key at ANTHROPIC_API_KEY, OPENAI_API_KEY, or GOOGLE_API_KEY, depending on which model you want to use. Claude 3.5 Sonnet is currently the best balance between quality and cost for most users.

Also set your timezone (TZ=Europe/Amsterdam), choose a strong password for the admin interface, and configure the messaging platforms you want to use. For WhatsApp you need a separate phone number; for Discord and Slack a bot token is sufficient. The documentation at docs.openclaw.dev describes every variable with examples.

A common mistake is forgetting to set the ENCRYPTION_KEY variable. OpenClaw encrypts all messages in the database by default. Generate a random key with openssl rand -hex 32 and paste it into your .env file.

Step 3: Start Containers and Health Check

Start OpenClaw with docker compose up -d. The first time, Docker downloads the required images — expect two to five minutes depending on your internet speed. Check that all containers are running with docker compose ps. You should see three services: openclaw-core, openclaw-web, and a PostgreSQL database.

Open your browser and navigate to http://localhost:3000 (or your domain name if you configured one). You will see the OpenClaw dashboard with a setup wizard. Log in with the admin password you set in step 2. The wizard automatically tests the connection to your AI provider and shows whether the messaging platforms are correctly configured.

Step 4: Connect Your First Messaging Platform

In the dashboard, go to Settings and then Channels. Choose the platform you use most. For Telegram the connection is simplest: create a bot via @BotFather in Telegram, copy the token, and paste it into OpenClaw. Within thirty seconds you can send a message to your AI assistant.

For WhatsApp the process is slightly more complex. OpenClaw uses the WhatsApp Business API via a self-hosted Baileys bridge. After scanning a QR code with your phone, the connection is active. Keep in mind that WhatsApp allows only one web connection per number — so use a separate number or accept that WhatsApp Web will be logged out.

Test the connection by sending a simple message like "What is the weather today?" If you get a coherent response, everything is working. Then try something more complex like "Summarize my last five emails" to see how OpenClaw handles tools and actions.

Step 5: Install Skills and Personalize

OpenClaw becomes more powerful with skills — plugins that add new functionality. In the dashboard, go to the Skill Store and browse the available options. Popular choices include the Calendar skill for Google Calendar integration, the Notes skill for task tracking, and the Web Search skill for real-time information.

Only install skills from the verified catalog (indicated by a blue checkmark). The community has experienced issues with malicious skills on unverified marketplaces in the past. After installing a skill, OpenClaw automatically restarts the necessary services — you do not need to manually restart anything.

Finally, personalize your assistant. Go to Settings and then Personality. Here you can specify how formal or informal OpenClaw should respond, in which language, and which topics should take priority. The more context you provide, the more useful the assistant becomes during the first few weeks.

Conclusion

With these five steps you have a fully working, self-hosted AI assistant running. The entire process takes between thirty and sixty minutes, depending on your experience with Docker and the complexity of your setup. The main advantage over cloud services is control: your data never leaves your own server, and you can customize the system exactly to your needs.

In future articles we will dive deeper into advanced configuration, including running local language models with Ollama, setting up automatic backups, and building your own custom skills.