OpenClaw Tutorial: Getting Started in 30 Minutes (2026)

Quick Answer: Install OpenClaw with npm install -g openclaw@latest, run openclaw onboard, and you'll have a working AI agent chatting with you in 30 minutes — for free, on your own machine.

OpenClaw crossed 250,000 GitHub stars in March 2026, making it one of the fastest-growing open-source projects in history. It did that by being the one personal AI agent that actually works the way you want: it lives on your machine, talks to you on the apps you already use, and can actually do things — browse the web, write files, run code — not just respond to prompts.

This tutorial walks you through the full setup: install, connect your first channel, add skills, and have a conversation with your new agent. No Docker required. No subscription. Just a terminal and 30 minutes.

What OpenClaw Actually Is

OpenClaw is a self-hosted agent runtime and message router. It sits in the background on your machine (or a server), connects to an LLM of your choice, and exposes that agent through the messaging channels you already use.

The analogy: imagine a very capable employee who lives in your home server room, reads every message you send through Slack or WhatsApp, and has access to your filesystem, browser, and terminal. That's OpenClaw.

Three things make it different from cloud assistants:

• Memory — it stores conversation history and personal context locally. Ask it a follow-up three days later and it remembers.
• Skills — 5,700+ plugins on ClawHub let you give it new abilities in seconds. Calendar, email, GitHub, PDF reading, home automation — all installable with one command.
• Channels — one agent, reachable from 20+ apps. Same brain, every surface.

Before You Start

You need three things before installing:

1. Node.js 20+ — check with node --version. Install from nodejs.org if needed.
2. An LLM API key — OpenClaw works with Anthropic (Claude), OpenAI, Google Gemini, DeepSeek, or any OpenRouter-compatible provider. For this tutorial we'll use Claude.
3. 10 minutes of focused time — the onboarding wizard handles everything else.

Step 1 — Install OpenClaw

Open a terminal and run:

npm install -g openclaw@latest

Or with pnpm:

pnpm add -g openclaw@latest

Verify the installation:

openclaw --version

You should see the version number printed. If you get command not found, make sure your npm global bin directory is on your $PATH.

Step 2 — Run the Onboarding Wizard

The onboarding wizard is the fastest way to get everything configured:

openclaw onboard

The wizard walks you through five steps:

1. Mode — QuickStart (recommended defaults) or Advanced (full control over every option)
2. Workspace — where to store your agent's data (default: ~/.openclaw)
3. Model provider — which LLM to connect (Claude, GPT-4o, Gemini, DeepSeek, local)
4. Channels — which messaging apps to enable
5. Skills — optional starter skills to install from ClawHub

For your first setup, choose QuickStart. It sets sensible defaults and skips options you can configure later.

Entering your API key

When the wizard asks for your AI provider, select Anthropic and paste your key from console.anthropic.com. The key is stored locally in ~/.openclaw/config.json — it never leaves your machine.

Installing as a daemon

If you want OpenClaw to start automatically when your machine boots, run:

openclaw onboard --install-daemon

This registers the Gateway as a launchd service on macOS or a systemd user service on Linux. You'll never need to manually start it again.

Step 3 — Start the Gateway

The Gateway is OpenClaw's brain — the server process that manages your agent, stores memory, and routes messages between channels.

If you installed as a daemon, it's already running. Otherwise, start it manually:

openclaw gateway

Leave this running in one terminal. Open a second terminal to interact with your agent.

Step 4 — Have Your First Conversation

The fastest way to test your agent is the built-in browser dashboard:

openclaw dashboard

This opens a web interface at http://localhost:4111 where you can chat with your agent right now — no channel setup needed.

Try it:

You: What can you do?
Agent: I can browse the web, read and write files on your machine, run shell
       commands, search for information, and help with tasks across the apps
       you connect me to. I also remember everything we've discussed...

Alternatively, chat from the terminal:

openclaw chat

Both work identically — it's the same agent, same memory, same skills.

Step 5 — Connect a Messaging Channel

This is where OpenClaw gets genuinely useful. Instead of opening a separate app, your agent meets you where you already are.

Connect Telegram (easiest)

1. Open Telegram and search for @BotFather
2. Send /newbot and follow the prompts to get a bot token
3. Run: openclaw channels add telegram
4. Paste your bot token when prompted

That's it. Message your new bot in Telegram and your OpenClaw agent responds.

Connect Slack

1. Create a Slack app at api.slack.com/apps
2. Enable Socket Mode and generate an app-level token
3. Add the bot to a channel and copy the OAuth token
4. Run: openclaw channels add slack
5. Paste both tokens when prompted

Your agent is now a member of your Slack workspace.

Connect via WhatsApp

openclaw channels add whatsapp

This generates a QR code. Scan it with your phone (WhatsApp > Linked Devices) to connect your WhatsApp account to the agent.

Step 6 — Add Skills from ClawHub

Skills extend what your agent can do. The ClawHub registry has 5,700+ skills covering almost every automation use case.

Search and install

# Search for available skills
npx clawhub search calendar

# Install a skill
npx clawhub install google-calendar

# Keep skills up to date
npx clawhub sync

Starter skills worth installing

After installing, your agent automatically knows how to use these tools. You don't need to configure prompts or write any code.

Auto-detected CLI tools

OpenClaw also scans for installed CLI tools and wraps them as skills automatically. If you have gh (GitHub CLI), ffmpeg, imagemagick, or yt-dlp installed, the agent can use them without any setup.

Step 7 — Give Your Agent Memory and Context

OpenClaw stores context in your local workspace (~/.openclaw). You can give it persistent facts about yourself:

openclaw context set "My name is Alex. I'm a backend engineer at a Series B startup.
I prefer concise answers. My timezone is UTC+3."

Now every conversation starts with that context already loaded. Ask it to schedule something and it knows your timezone. Ask for a code review and it knows your background.

You can also ask the agent to remember things mid-conversation:

You: Remember that the production database uses PostgreSQL 15.
Agent: Got it — I've saved that to your workspace context.

Useful Commands Reference

Troubleshooting Common Issues

Gateway won't start: Check that port 4111 is free (lsof -i :4111). Change the port in ~/.openclaw/config.json if needed.

Agent doesn't respond in Telegram: Make sure the Gateway is running and the bot token is correct. Run openclaw channels list to verify the channel shows as "connected."

"No tool found" errors: A required skill isn't installed. Run npx clawhub search <task> to find the right one and install it.

API key rejected: Double-check the key hasn't expired or been rotated. Edit ~/.openclaw/config.json to update it.

OpenClaw vs. Managed Alternatives

OpenClaw gives you maximum control and zero ongoing cost — but you're responsible for keeping the Gateway running and the software updated.

If you want the same capabilities without the self-hosting overhead, a managed alternative makes sense:

For solo developers and privacy-conscious users, OpenClaw is the better fit. For teams who want a shared agent without anyone running a server, cowork.ink gives everyone access to the same AI agent and context — no Gateway to babysit.

What to Do Next

Your agent is running. Here's where to go from here:

1. Install 3–5 skills from ClawHub that match your daily work (calendar, GitHub, email are popular starting points)
2. Set your personal context with openclaw context set — the more you tell it, the more useful it gets
3. Connect your primary messaging channel so you stop switching apps
4. Explore the dashboard at localhost:4111 to see memory, skill logs, and conversation history

The agent gets dramatically more useful after a week of use as it builds context about how you work. The lobster 🦞 takes a little while to learn your habits — give it time.

Get Started with OpenClaw

Install it now:

npm install -g openclaw@latest && openclaw onboard

For the full reference, visit docs.openclaw.ai.

If you're on a team and want everyone to share the same AI agent without anyone running a server, cowork.ink is the zero-ops path — shared workspace, shared context, no Gateway to manage.

OpenClaw was created by Peter Steinberger and is now maintained by the OpenClaw Foundation under an MIT license. It was formerly known as Clawdbot and Moltbot.

Frequently Asked Questions

What is OpenClaw and is it free?

OpenClaw is a free, open-source personal AI agent you run on your own machine. It connects to any LLM provider (Claude, GPT-4o, Gemini, DeepSeek, or a local model) using your own API key — no subscription needed. The software itself is MIT-licensed.

What messaging apps does OpenClaw support?

OpenClaw works with WhatsApp, Telegram, Slack, Discord, iMessage (via BlueBubbles), Signal, Google Chat, Microsoft Teams, Matrix, and many others. You can talk to the same agent from multiple channels simultaneously.

Do I need a powerful computer to run OpenClaw?

No. OpenClaw's Gateway (the agent server) is lightweight — it runs comfortably on a Raspberry Pi, an old laptop, or a $5/month VPS. The heavy lifting is done by the cloud LLM you connect it to, not your hardware.

How do I add new capabilities to OpenClaw?

Install Skills from ClawHub, the community registry with 5,700+ plugins. Run `npx clawhub install <name>` to add any skill. OpenClaw can also auto-detect installed CLI tools (like `gh` or `ffmpeg`) and wrap them as skills automatically.

What is the difference between OpenClaw and a chatbot like ChatGPT?

ChatGPT runs in the cloud and forgets each conversation. OpenClaw runs on your hardware, retains memory across all sessions, can take action (browse the web, read and write files, run shell commands), and receives your messages through the apps you already use. It's an agent, not a chat widget. See our [comparison of AI agents vs. chatbots](/blog/ai-agents-vs-chatbots/) for a full breakdown.