Lesson 4. First Launch and Setup#

Why this lesson?#

OpenClaw is installed — now you need to set it up. You’ll connect an AI model (Claude or GPT), start the Gateway, and have your first conversation with the assistant. Sounds complicated? It’s actually not: OpenClaw has a setup wizard that will guide you step by step.

Step 1. Launch the setup wizard#

Open a terminal and type:

openclaw onboard --install-daemon

What this command means:

• openclaw onboard — launches the setup wizard. It will ask you questions and configure everything automatically.
• --install-daemon — also installs OpenClaw as a background service (daemon), so it keeps running even after you restart your computer.

Daemon is a program that runs in the background, like an alarm on your phone: you don’t see it, but it’s always ready to go off.

Step 2. Connect an AI model#

The wizard will ask which AI service you want to use. There are two main options:

Option A: Anthropic (Claude) — recommended#

Claude is an AI model from the company Anthropic. OpenClaw developers recommend it (especially Claude Pro/Max models) for its quality and safety.

How to connect:

1. The wizard will offer to sign in via OAuth (authorization through the browser)
2. An Anthropic page will open — sign in to your account or create a new one
3. Confirm access
4. Done — OpenClaw will have a connection to Claude

OAuth is a sign-in method where you allow one program to use your account in another. Like “Sign in with Google” on various websites. You don’t share your password — you just grant access.

Option B: OpenAI (GPT)#

If you prefer ChatGPT:

1. The wizard will also offer OAuth authorization
2. Sign in to your OpenAI account
3. Confirm access

Subscription: you need a paid subscription to one of the services (Anthropic Pro/Max or ChatGPT Plus/Pro). Free accounts usually have significant limitations.

Step 3. Gateway setup#

The wizard will also set up the Gateway — the “heart” of OpenClaw, which:

• Receives your messages from messengers
• Sends them to the AI model
• Returns responses to you

Usually everything is configured automatically. The Gateway runs on port 18789 (this is like a “room number” inside your computer).

Step 4. Check that everything works#

After the wizard finishes, let’s check the status:

openclaw gateway status

You should see a message that the Gateway is running. ✅

Step 5. Your first conversation!#

Now let’s open the Dashboard — a web interface where you can chat with AI right in your browser:

openclaw dashboard

A browser will open with an address like http://127.0.0.1:18789/. You’ll see a chat — write something!

Try:

• “Hello! Tell me about yourself”
• “What’s the weather like today?”
• “Write a short poem about a cat”

If you got a response — everything works! 🎉

Alternative: chat through the terminal#

You can talk to the assistant right in the terminal:

openclaw agent --message "Hello! Who are you?"

If something went wrong#

Run the diagnostics:

openclaw doctor

This command will check all settings and tell you what needs to be fixed.

Useful commands#

Lesson summary#

• The setup wizard openclaw onboard does most of the work for you
• You connected an AI model (Claude or GPT) via OAuth
• Gateway is a background service that connects you to AI
• You can have your first conversation through the Dashboard (web interface) or through the terminal
• If something doesn’t work — openclaw doctor will help find the problem

In the next lesson, we’ll connect a messenger so you can chat with AI right in Telegram or WhatsApp!