Getting Started with OpenClaw — install, onboard and chat in 5 minutes

Beginner installonboardingclibeginner OpenClaw Guide Editorial 📅 2026-09-21 ⏱ 6 min

A first-time walkthrough for installing the open-source OpenClaw AI assistant, running the onboarding wizard and holding your first conversation.

OpenClaw is an open-source AI assistant that runs entirely on your machine. It is steered by the OpenClaw Foundation, an independent 501(c)(3), and its tagline is honest: "No subscription. No hosted tier. No token. Your memories. Your skills. Your models. Your machines. Your data." This tutorial gets you from zero to a working assistant in about five minutes.

What you need

A computer running macOS 15+, Windows 10/11 (x64 or ARM64), or Linux x64.
About 500 MB of free disk space for the runtime, plus whatever your model provider needs.
An account or API key with a model provider (OpenAI, Anthropic, Google, OpenRouter, local Ollama, etc.).

The desktop apps bundle Node.js 24.16+ or 26.1+ for you, so you do not need to install it yourself unless you want to.

OpenClaw is not a hosted service. The Gateway runs as a background process on your own machine and exposes a local dashboard. Your memories, your skills and your chats stay on your hardware.

Install with the desktop app (recommended)

Download the installer for your platform from the official install page. The current stable release is v2026.9.5 for macOS and v2026.9.4 for Windows.
Move the app to /Applications (macOS) or run the installer (Windows). The app installs the Gateway, the chat client and the setup wizard in a single step.
Open OpenClaw and follow the onboarding wizard. You will be asked to:

Pick a model provider.
Sign in or paste an API key.
Let the wizard install the Gateway as a background service.
Optionally connect a chat app (Telegram, Discord, Slack, WhatsApp, iMessage).

When onboarding finishes you land in the chat dashboard. Type a question to confirm everything is wired up:

hello! who are you and what can you do?

Install with the one-liner (terminal)

If you prefer a clean install on a server or a headless box, the terminal installer is faster than the GUI:

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

The installer downloads Node.js if needed, then starts the onboarding wizard. On macOS the first run may ask for an Administrator password to install Homebrew.

To install the CLI without the onboarding prompts (for scripted deployments):

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

This keeps Node.js and OpenClaw under a local prefix such as ~/.openclaw and changes nothing else on the system.

npm / pnpm / Bun

If you already manage Node yourself:

npm i -g openclaw
openclaw onboard

On npm 12+ pass --allow-scripts=openclaw so the install hooks can run.

pnpm add -g --allow-build=openclaw openclaw@latest
openclaw onboard

What the Gateway actually is

OpenClaw ships a small Gateway service that runs in the background. It owns:

The local HTTP API on http://127.0.0.1:18789 that the chat UI, skills and SDKs talk to.
The channel adapters that bridge Telegram, Discord, Slack, WhatsApp and iMessage.
The memory store that persists your long-term facts.

You do not interact with the Gateway directly — the chat UI does — but it is worth knowing it is there because every command below manipulates it.

Your first chat from the CLI

Open a terminal and run:

openclaw chat "draft a polite follow-up email to a client I met last week"

The first call can take a few seconds while the model provider responds. You will see the assistant stream tokens back, then a final message you can paste anywhere.

To open a persistent interactive session, simply run:

openclaw chat

Type exit or press Ctrl-D to leave. The conversation is saved to your local memory and is searchable from the dashboard.

Verify the install

The doctor command diagnoses the most common setup problems:

openclaw doctor

A healthy install prints a green ✓ for every check. A red ✗ tells you exactly which dependency is missing.

Update and switch channels

OpenClaw ships on a stable channel by default. To switch to the next beta release early:

openclaw update --channel dev

To come back to stable:

openclaw update --channel stable

Next steps

Connect Telegram, WhatsApp or Discord so you can message your assistant from the chat apps you already use.
Learn how to build reusable skills so the assistant can call APIs, run shell commands or read your calendar.
Set up long-term memory so facts about you survive across sessions.
When something breaks, head to Troubleshooting.

OpenClaw Guide Editorial

Independently maintained walkthroughs and ecosystem notes for OpenClaw.

Found an issue or have a correction? Get in touch.