How to Install OpenClaw: Step-by-Step Setup Guide (2026)

Get OpenClaw running on your machine in under 20 minutes. Four installation methods, model configuration, and channel connections for Telegram, WhatsApp, and Discord.

Installing OpenClaw takes under 20 minutes. Run the one-line installer, connect an AI model (Anthropic, OpenAI, or a free local model via Ollama), plug in a messaging channel, and you're live. OpenClaw is free, open-source software -- the only cost is the AI API usage behind it.

This guide covers four installation methods, the onboarding wizard, model setup, and connecting your first channel. Pick the method that fits your comfort level.

Jump to: Requirements · One-Line Installer · npm Install · Docker · Build from Source · Onboarding Wizard · AI Model Setup · Connect a Channel · Verify · Troubleshooting

System Requirements

Before you start, make sure your machine meets these minimums:

Method 1: One-Line Installer (Recommended)

The fastest path. One command installs OpenClaw, its background daemon, and all dependencies.

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

wsl --install

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

After installation, verify it worked:

openclaw --version

You should see something like openclaw v1.x.x.

Method 2: npm Install

If you already have Node.js 22+ and prefer managing packages through npm:

npm install -g openclaw@latest

This installs the openclaw CLI globally. Verify:

openclaw --version

Updating via npm

Unlike the one-line installer (which auto-updates), npm installs need manual updates:

npm update -g openclaw

Method 3: Docker

Docker is ideal if you want isolation, reproducibility, or plan to deploy OpenClaw on a server. You need Docker Engine 24+ and Docker Compose v2.

mkdir openclaw && cd openclaw

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

docker compose up -d

docker compose logs -f openclaw

Method 4: Build from Source

For contributors or anyone who wants the absolute latest (possibly unstable) code:

git clone https://github.com/openclaw-ai/openclaw.git
cd openclaw
npm install
npm run build
npm link

This makes the openclaw command available globally via symlink. To update, pull the latest and rebuild:

git pull && npm install && npm run build

The Onboarding Wizard

After installation, the onboarding wizard walks you through first-time setup. It configures your AI model, installs the background daemon, and optionally connects your first channel.

openclaw onboard --install-daemon

The --install-daemon flag sets up the background process that keeps OpenClaw running even when you close your terminal. The wizard will ask you three things:

1. AI model provider -- Which API to use (Anthropic, OpenAI, OpenRouter, or Ollama for local models)
2. API key -- Your key for the chosen provider
3. Channel connection -- Optionally connect Telegram, WhatsApp, or Discord right away

You can skip the channel step and add channels later with openclaw channel add.

Setting Up Your AI Model

OpenClaw doesn't include an AI model -- it connects to one. You have three options, from highest quality to lowest cost:

Option A: Anthropic API (Recommended)

Best quality. Get your API key from console.anthropic.com.

openclaw config set model.provider anthropic
openclaw config set model.apiKey sk-ant-your-key-here
openclaw config set model.name claude-sonnet-4-5

Cost: roughly $3 per million input tokens, $15 per million output tokens with Sonnet 4.5. For most users that's $10-30/month. You can also use Haiku 4.5 ($1/$5) for lighter tasks or Opus 4.6 ($5/$25) for maximum quality.

Option B: OpenRouter (Multi-Model Access)

OpenRouter gives you one API key for dozens of models -- Claude, GPT-4o, Llama, Mistral, and more. Useful if you want to experiment or switch models without managing multiple accounts.

openclaw config set model.provider openrouter
openclaw config set model.apiKey sk-or-your-key-here
openclaw config set model.name anthropic/claude-sonnet-4.5

Sign up at openrouter.ai. Pricing varies by model but is typically close to the original provider's rates.

Option C: Ollama (Free, Local Models)

Run AI models entirely on your machine with zero API costs. Quality is lower than cloud models, but it's completely free and private.

# Install Ollama first
curl -fsSL https://ollama.com/install.sh | sh

# Pull a model (Llama 3.3 8B is a good starting point)
ollama pull llama3.3

# Configure OpenClaw to use it
openclaw config set model.provider ollama
openclaw config set model.name llama3.3

No API key needed. Ollama runs locally on port 11434. You'll need at least 8 GB of RAM for the 8B model, or 16 GB for the 70B model.

Connecting Your First Channel

OpenClaw supports Telegram, WhatsApp, and Discord out of the box. You only need one to get started.

Telegram (Easiest)

1. Open Telegram and search for @BotFather
2. Send /newbot and follow the prompts to name your bot
3. BotFather gives you a token like 123456789:ABCdefGHIjklMNO
4. Add it to OpenClaw:

openclaw channel add telegram --token "YOUR_BOTFATHER_TOKEN"

Your bot is now live. Send it a message in Telegram to test.

WhatsApp

1. Run the WhatsApp connector:

openclaw channel add whatsapp

2. A QR code appears in your terminal
3. Open WhatsApp on your phone, go to Linked Devices, and scan the QR code
4. Done. Messages to the linked number now route through OpenClaw

Discord

1. Go to the Discord Developer Portal
2. Create a new application, then create a Bot under the Bot tab
3. Copy the bot token
4. Under OAuth2, generate an invite URL with bot and applications.commands scopes
5. Invite the bot to your server, then:

openclaw channel add discord --token "YOUR_DISCORD_BOT_TOKEN"

Verify Your Installation

OpenClaw includes a built-in health checker. Run:

openclaw doctor

This checks Node.js version, daemon status, model connectivity, and channel health. You'll see output like:

OpenClaw Doctor v1.4.2
------------------------------
Node.js ............ v22.11.0  OK
Daemon ............. running   OK
Model provider ..... anthropic OK
  Model ............ claude-sonnet-4-5 OK
  API key .......... sk-ant-***  OK
Channels:
  telegram ......... connected OK
------------------------------
All checks passed.

To check just the gateway (channel connections):

openclaw gateway status

This shows each connected channel, its status, and the last message timestamp.

Troubleshooting Common Issues

"command not found: openclaw"

The npm global bin directory isn't in your PATH. Fix it:

export PATH="$PATH:$(npm config get prefix)/bin"

Make it permanent by adding that line to ~/.zshrc (Mac) or ~/.bashrc (Linux/WSL).

Node.js version too old

OpenClaw requires Node.js 22+. Check your version:

node --version

If it shows v18 or v20, upgrade. The easiest way:

# Mac
brew install node@22

# Linux/WSL
curl -fsSL https://deb.nodesource.com/setup_22.x | sudo -E bash -
sudo apt-get install -y nodejs

PORT 3000 already in use

Another process is using the default OpenClaw port. Either stop it or change OpenClaw's port:

# Find what's using port 3000
lsof -i :3000

# Or change OpenClaw's port
openclaw config set server.port 3001

Docker: Image not found for ARM

If you're on Apple Silicon or another ARM machine and get an image architecture error, make sure you're pulling the multi-arch image:

docker pull --platform linux/arm64 openclaw/openclaw:latest

The official image supports both amd64 and arm64. If you're using an older Docker version, update to Docker Engine 24+.

Daemon won't start

Check the daemon logs:

openclaw daemon logs

Common causes: port conflict, missing API key, or insufficient permissions. If the daemon crashes on startup, try reinstalling:

openclaw daemon uninstall
openclaw onboard --install-daemon

WhatsApp QR code expired

QR codes expire after 60 seconds. Run openclaw channel add whatsapp again for a fresh one. Make sure your phone and computer are on the same network.

What's Next?

Now that OpenClaw is installed and connected, here's where to go from here:

• What Is OpenClaw? -- understand the full feature set
• OpenClaw Pricing Guide -- understand and control your AI costs
• WhatsApp & Telegram Setup -- advanced channel configuration
• How to Install Claude Code -- set up Anthropic's coding agent
• OpenClaw vs Claude Code -- compare both AI agents