Setup Guide

Zero to running in 5 minutes

A complete, no-gaps guide from install to your first autonomous task. Every step is tested and documented.

1

Prerequisites

Node.js (Required)

OpenClaw requires Node.js version 22.14 or higher. Node.js v24 is recommended for best compatibility and performance.

⚠️
Version matters Older Node.js versions cause silent failures during Gateway startup. Always verify before installing OpenClaw.
Check your Node version
$node --version
# Should show v22.14.0 or higher
v24.0.0

If Node.js is not installed or your version is too old, install it:

macOS — Homebrew
$brew install node@24
# Or download from nodejs.org
Windows — PowerShell
# Download installer from nodejs.org/en/download
# Or use winget:
>winget install OpenJS.NodeJS.LTS
Linux — apt
$curl -fsSL https://deb.nodesource.com/setup_24.x | sudo -E bash -
$sudo apt-get install -y nodejs
Any OS — nvm (recommended)
$nvm install 24
$nvm use 24
Now using node v24.0.0

API Key for your AI model

You'll need an API key from at least one LLM provider. OpenClaw is model-agnostic — use whichever model you prefer.

🤖 Anthropic Claude
console.anthropic.com
🧠 OpenAI GPT-4o
platform.openai.com
🔵 DeepSeek
platform.deepseek.com
🦙 Ollama (local)
ollama.com — free & private
🔐
Security first Create dedicated bot/service accounts before running OpenClaw. Never connect it to your primary personal Telegram, WhatsApp, or GitHub account. OpenClaw can execute shell commands and access files — isolate it.
2

Installation

Run the install script

OpenClaw provides a one-line installer for all platforms. This downloads the binary, sets up the CLI, and prepares your environment.

macOS / Linux
$curl -fsSL https://openclaw.ai/install.sh | bash
✓ OpenClaw installed successfully
✓ CLI available: openclaw --version
Windows PowerShell (run as Admin)
>iwr -useb https://openclaw.ai/install.ps1 | iex
✓ OpenClaw installed successfully
ℹ️
Verify installation After installation, open a new terminal and run openclaw --version. If the command is not found, restart your terminal or add the install path to your PATH.
3

Onboarding Wizard

Start the interactive setup

The --install-daemon flag also registers OpenClaw as a background service so it starts automatically on reboot.

Start onboarding
$openclaw onboard --install-daemon
✓ Node.js v24.0.0 detected
✓ Gateway initialized on port 18789
✓ Daemon registered
→ Let's configure your AI model...
💡
What the wizard does It walks you through: selecting your AI provider, entering your API key, creating a workspace, and pairing your first messaging channel. Answer the prompts step by step.
4

Connect Your AI Model

Choose your provider

During onboarding, the wizard will prompt you to select an AI model provider. Your API key is stored locally and never sent to OpenClaw servers.

Model configuration prompt
? Select your AI model provider:
❯ Anthropic (Claude)
OpenAI (GPT-4o)
DeepSeek
Ollama (local model)
Custom provider
? Enter your Anthropic API key: sk-ant-••••••••
✓ API key validated and stored securely
🔒
Custom provider support If your model isn't listed, choose "Custom provider" and enter your API base URL, API key, and model ID. Works with any OpenAI-compatible API endpoint.
5

Connect a Messaging Channel

Telegram (Recommended for beginners)

Telegram is the easiest channel to set up and has the most reliable integration. Start here if unsure.

  1. Open Telegram and search for @BotFather
  2. Send /newbot and follow the prompts to create a new bot
  3. Copy the bot token provided (format: 1234567890:ABCdef...)
  4. Paste the token when OpenClaw's wizard asks for it
Telegram pairing
? Select a channel to connect: Telegram
? Enter your bot token: 1234567890:ABCdef...
✓ Telegram bot connected
✓ Pairing code sent to your bot chat
? Enter pairing code from Telegram: ••••
✓ Channel authorized. Send a message to start!

WhatsApp

WhatsApp uses a QR code pairing flow. You'll need to scan it from your WhatsApp app on your phone.

⚠️
Use a dedicated WhatsApp number Do not connect your primary personal number. Create a new dedicated number (even a cheap SIM) for your OpenClaw agent to avoid conflicts and protect your data.

Discord

Discord requires setting up a bot in the Discord Developer Portal. There's one critical setting you must enable.

  1. Go to discord.com/developers/applications
  2. Create a new application → Bot → copy the bot token
  3. Under "Privileged Gateway Intents" enable Message Content Intent
  4. Invite the bot to your server with Send Messages + Read Messages permissions
🚫
Message Content Intent is mandatory Forgetting to enable this causes the bot to receive zero messages — no errors thrown. This is the #1 Discord setup mistake.
6

Verify Your Setup

Check Gateway status

The Gateway should be listening on port 18789. If it's not running, your agent won't respond.

Verify gateway
$openclaw gateway status
✓ Gateway: RUNNING
✓ Listening on port 18789
✓ Connected channels: Telegram [mybotname]
✓ AI model: Anthropic claude-3-5-sonnet
Health check
$openclaw doctor
✓ Node.js version: OK
✓ Gateway: running
✓ API key: valid
✓ Channels: 1 connected
✓ Workspace: valid
✓ No configuration issues found

Open the Dashboard

OpenClaw includes a browser-based control panel for managing sessions, viewing logs, and adjusting configuration.

$openclaw dashboard
✓ Dashboard opened at http://localhost:18789/ui
7

Configure Your Workspace & Skills

Workspace structure

Your agent's workspace is a local folder with configuration files. Edit these to customize your agent's persona, behavior, and capabilities.

workspace/
├── SOUL.md       — Agent persona, instructions, tone
├── AGENTS.md     — Agent config, model settings, DM policies
└── skills/       — Custom capability definitions
    ├── research.md
    ├── outreach.md
    └── dev-ops.md

Creating your first skill

Skills are Markdown files with YAML frontmatter that describe a capability to the agent. Create one in workspace/skills/.

Example skill: competitor-research.md
---
name: Competitor Research
description: Researches competitors and returns structured analysis
tools: [web_search, web_browser, file_write]
---
# Instructions
When asked to research competitors, search Product Hunt,
G2, and Crunchbase. Summarize top 5 competitors with
pricing, features, and identified gaps. Save as PDF.
8

Always-On Setup (24/7)

Run on a dedicated machine

For a true always-on assistant, run OpenClaw on a machine that never sleeps. Popular options:

🖥️ Mac Mini
Ideal for local, low power, always running. M4 recommended.
☁️ VPS / Cloud
DigitalOcean, Hetzner, or Contabo. $5–10/month.
🏠 Home Server
Raspberry Pi or old laptop. Perfect for self-hosting.
⚠️
Don't run on your daily laptop Your agent stops when you close the lid. For 24/7 operation, use a dedicated machine or a VPS.
✓

Run Your First Task

🚀 You're live. Start delegating.

Open your connected messaging app (Telegram, WhatsApp, or Discord) and send your agent a message. Try something simple first:

Hello! What can you do?
OpenClaw Hi! I'm your personal AI assistant. I can research topics, manage files, run terminal commands, browse the web, call APIs, and execute custom workflows. What would you like to delegate first?
Research the top 5 no-code tools for founders and summarize them
OpenClaw On it. Searching Product Hunt, G2, and 3 tech newsletters...
✓ Done. Summary saved to your workspace. Here's the quick overview...
🎉
Setup complete You have a working autonomous AI agent. From here, explore the Docs to add more skills, connect more channels, and build real workflows.
Explore Use Cases → View Full Docs FAQ