The fastest way to get started with OpenClaw

Two things. A Mac, a Linux machine or Windows. And access to a model: an API key from a provider, or an existing Claude Code or Codex login, which onboarding can reuse.

You do not need Node.js installed. The install script brings a supported version if yours is missing. If you would rather use npm, you need Node 24.22.1 or later, or 26.12.1 or later.

Installed with npm instead? Then start the wizard yourself with openclaw onboard --install-daemon. The daemon part keeps the gateway running in the background after you close the terminal.

The dashboard is only the start. Most people want to message their assistant from their phone, so the next step is a channel. OpenClaw's channel guides cover WhatsApp, Telegram, Slack, Discord, iMessage and more than twenty others.

When a stranger messages your assistant, OpenClaw does not answer them. It holds them as a pairing request until you approve it with openclaw pairing approve. Keep it that way. The docs also say tools run on your machine, not in a sandbox, unless you set one up. Read the security guide before you add other people.

OpenClaw runs as a gateway on the machine you installed it on. Its own FAQ is plain about what that means. When a laptop sleeps or loses its network, the gateway disconnects, and updates and reboots interrupt it. For an assistant you message at night, that is most of the day.

The FAQ's answer is a dedicated host. A small server with one CPU and 1 GB of memory is the minimum; 2 GB is what it recommends. You then look after that server: updates, a firewall, and the gateway itself.

Or skip the server. OpenClaw on Faivelo runs on a box of its own with 1 CPU, 2 GB of memory and 20 GB of disk, for $9 a month. You pick a provider and paste your key when you set it up, and it is checked as you paste it. It gets its own email address, so you can write to it like a person. You link WhatsApp, Telegram, Discord or Slack from its page. Weighing it against Hermes first? See OpenClaw vs Hermes Agent.