OpenClaw is one of the fastest-growing AI agent platforms right now — and for good reason. It turns any machine into a self-hosted AI gateway that you can message from WhatsApp, Telegram, Discord, iMessage, and more. One install, every channel, your data stays on your hardware.

This is the complete, accurate setup guide for February 2026. Follow it top to bottom and you’ll have a working personal AI agent in under 15 minutes.

—

What Is OpenClaw?

OpenClaw is a self-hosted AI agent gateway. You install it on your own machine (laptop, desktop, or server), connect it to your AI provider of choice (Anthropic, OpenAI, Google, or even a local Ollama model), and then message it through the chat apps you already use.

Key things that make it different from ChatGPT or Claude.ai:

Multi-channel: one gateway serves WhatsApp, Telegram, Discord, iMessage — simultaneously
Self-hosted: your conversations and files never leave your machine
Agent-native: persistent memory, tool use, skills, cron jobs, and multi-agent routing built in
Open source: MIT licensed, actively maintained

—

System Requirements

Node.js 22+ (the installer will handle this if you don’t have it)
macOS, Linux, or Windows (WSL2 recommended on Windows)
An API key from Anthropic, OpenAI, or another supported provider

That’s it.

—

Step 1: Install OpenClaw

The recommended method is the one-line installer, which handles Node detection, downloads the CLI, and launches the setup wizard automatically.

macOS / Linux / WSL2:

curl -fsSL https://openclaw.ai/install.sh | bash

Windows (PowerShell):

iwr -useb https://openclaw.ai/install.ps1 | iex

Prefer npm? That works too:

npm install -g openclaw@latest

—

Step 2: Run the Onboarding Wizard

After install, run:

openclaw onboard --install-daemon

The wizard walks you through:

Connecting your AI provider (paste your API key)
Choosing a model (Claude Sonnet is a strong default)
Installing the Gateway as a background service (daemon)
Optional: connecting your first channel (WhatsApp, Telegram, etc.)

The --install-daemon flag makes OpenClaw start automatically on boot — highly recommended so your agent is always available.

—

Step 3: Verify It’s Running

openclaw gateway status

You should see the Gateway reported as running. If anything looks off:

openclaw doctor

This runs diagnostics and tells you exactly what’s wrong and how to fix it.

—

Step 4: Open the Control UI

openclaw dashboard

This opens the browser-based Control UI at http://127.0.0.1:18789/. You can chat with your agent here immediately — no channel setup required.

If it loads, you’re done. Your personal AI agent is live.

—

Step 5: Connect a Channel (Optional but Recommended)

The real power of OpenClaw comes from messaging your agent through apps you already use. Connect a channel with:

openclaw channels login

Supported channels include:

WhatsApp — scans a QR code via WhatsApp Web
Telegram — connect via a bot token from BotFather
Discord — connect a Discord bot to your server
iMessage — macOS only, via local imsg CLI
Signal, Slack, Google Chat, MS Teams, and more

Once connected, you can message your agent from your phone just like any contact.

—

Step 6: Configure Who Can Access It

OpenClaw defaults to safe settings, but you’ll want to lock down who can message your agent. In ~/.openclaw/openclaw.json:

{
  "channels": {
    "whatsapp": {
      "allowFrom": ["+15555550123"]
    },
    "telegram": {
      "dmPolicy": "pairing"
    }
  }
}

allowFrom restricts to specific phone numbers. dmPolicy: "pairing" means only users you’ve explicitly paired can interact with your agent.

The Gateway watches the config file and reloads changes automatically — no restart needed.

—

Troubleshooting

openclaw not found after install:
Your npm global bin directory isn’t in your PATH. Fix:

export PATH="$(npm prefix -g)/bin:$PATH"

Add that line to your ~/.zshrc or ~/.bashrc.

Gateway won’t start:
Run openclaw doctor --fix — it identifies and patches the most common config issues automatically.

Channel won’t connect:
Check channel-specific docs at openclaw.ai/docs — WhatsApp and iMessage have platform-specific quirks.

—

What’s Next?

Once you’re up and running:

Set up skills — add capabilities to your agent (web search, weather, image generation, etc.)
Configure memory — teach your agent to remember things about you across sessions
Set up cron jobs — have your agent proactively check your email, calendar, or anything else on a schedule
Add a second agent — route different channels to different personas with openclaw agents add

—

Questions about your specific setup? Drop them in the comments — OS, install method, channel — and let’s get you sorted. 👇