How to Set Up OpenClaw in 15 Minutes

Quick Answer

Install Node.js 20+, then run npm install -g openclaw@latest followed by openclaw onboard --install-daemon. The guided wizard handles the rest — choosing a channel, connecting your model API key, and getting the daemon running. You can start chatting with your agent at http://127.0.0.1:18789/ in minutes.

Before You Start

You'll need three things: Node.js version 20 or higher (download from nodejs.org if you don't have it), a terminal or command prompt open, and an API key for the model you want to use. Most people start with Claude via Anthropic or GPT-4 via OpenAI — both offer free trial credits to get going.

That's it. No Docker, no cloud account, no config files to manually edit. The wizard handles everything else.

Step 1: Install OpenClaw

Open your terminal and run:

npm install -g openclaw@latest

This installs the OpenClaw command-line tool globally on your machine. It takes a minute or two depending on your connection speed.

Step 2: Run the Onboarding Wizard

Once installed, run:

openclaw onboard --install-daemon

The --install-daemon flag sets OpenClaw up as a background process so it keeps running even when you close your terminal. The wizard walks you through setting up your workspace and agent name, choosing messaging channels to connect (you can skip this for now), entering your model API key, and optionally installing starter skills.

When the wizard asks something you're unsure about, the default option is almost always fine for a first setup. Don't overthink it.

Step 3: Open the Control UI

Once onboarding finishes, open a browser and go to:

http://127.0.0.1:18789/

This is your Control UI — a simple dashboard where you can chat with your agent, manage skills, and see what's happening. You don't need to connect WhatsApp or Telegram to start experimenting. The Control UI is a perfectly good place to test things out.

Step 4: Say Hello

Type something in the chat window. Ask your agent to search for something, summarise a web page, or just introduce itself. If you get a response, the setup is working. Your agent is live.

Step 5: Connect Your Phone (Optional but Recommended)

One of the best things about OpenClaw is being able to message your agent from your phone, anywhere, like you would a contact in WhatsApp. To set this up, go to Settings → Channels in the Control UI.

For Telegram, you create a bot via BotFather (a five-minute process) and paste the token in. For WhatsApp, you scan a QR code. Either way it takes about five minutes and the instructions in the Control UI are clear.

Step 6: Install Your First Skill

Out of the box your agent can search the web and read files. Skills extend this further. A good first install is Summarize, which lets your agent condense any article, video, or document you send it:

npm i -g clawhub

clawhub install summarize

After installing, restart your agent and try sending it a link to an article. It should return a clean summary. That's your first real taste of what having an agent actually feels like.

Running It in the Cloud

Running OpenClaw locally is great for experimenting, but your agent goes offline when your computer does. If you want it running around the clock — so scheduled tasks and monitoring automations keep working while you sleep — the next step is deploying it to a cloud server or using Tulip, which handles the infrastructure for you.

Frequently Asked Questions

What if the install fails?

The most common issue is the wrong Node.js version. Check yours with node --version — you need 20 or higher. If you're on an older version, update via nodejs.org or use a version manager like nvm.

Does this work on Windows?

Yes. OpenClaw runs on Windows, macOS, and Linux. The setup process is identical on all three.

Is my data safe when running locally?

Everything stays on your machine. The only data that leaves is your prompts and responses going to whatever model API you configured. If you want full local privacy with no external API calls at all, OpenClaw supports running with a local model.

How do I stop or restart the daemon?

Run openclaw stop to stop it and openclaw start to restart. The Control UI also has start and stop controls if you prefer clicking over typing.

Can I have more than one agent?

Yes. You can create multiple agents in OpenClaw, each with a different name, skill set, and channel. A common pattern is a personal agent on one channel and a more focused work agent on another.