Want to try OpenClaw quickly but not sure where to start? This guide covers pre-install prep (Node version and port 18789), three installation options and 10-minute routes, steps from install to first chat, and three common pitfalls when the console is unreachable, plus FAQ, so you can get your first conversation running in about 10 minutes on Windows, Linux, macOS, or a remote Mac.
Pre-install: Node version and port 18789
OpenClaw 2026 requires Node.js v22 or higher; older versions can cause install or build failures. Prefer nvm: nvm install 22 && nvm use 22. OpenClaw uses port 18789 for the web console by default. On a cloud server or remote Mac, you must allow 18789 (TCP) in the firewall/security group, or the browser cannot reach the console.
- Node version: Run
node -vin the terminal; must be ≥ v22.0.0. - Port 18789: Allow inbound 18789 on the host; on Linux with ufw:
ufw allow 18789.
Three installation options and 10-minute route choice
Depending on your environment and preference, choose one of the three options below; each aims to get you installed and the console open in about 10 minutes.
| Method | Best for | Typical time | Notes |
|---|---|---|---|
| Official install script | Beginners, local or VPS quick try | ~5–10 min | curl -fsSL https://openclaw.ai/install.sh | bash; script checks and installs Node. |
| npm global install | Developers who already have Node 22+ | ~5–8 min | npm install -g openclaw@latest, then openclaw onboard and openclaw gateway. |
| Docker / pre-built image | Teams or production, isolated dependencies | ~10–15 min | Use official or community Docker image; map port 18789. Some clouds offer one-click OpenClaw images. |
npm config set registry https://registry.npmmirror.com.Steps from install to first chat
Using the install script and onboard, the minimal path is as follows.
Install OpenClaw
Run curl -fsSL https://openclaw.ai/install.sh | bash or npm install -g openclaw@latest and wait for completion.
Run setup wizard
Run openclaw onboard and follow prompts for Gateway, API, etc.; if a permission dialog appears (e.g. on macOS), allow it in the GUI.
Start Gateway
Run openclaw gateway (or use systemd/pm2 for persistence); ensure port 18789 is listening.
Open web console
In a browser go to http://localhost:18789 (or your server IP:18789) to open the Dashboard.
Send your first command
In the console or TUI (openclaw tui) send a simple command to confirm the model and Gateway respond; that completes your “first chat”.
Three pitfalls: port and console unreachable
New users often get stuck on “installed but cannot open the console” or errors during onboard. Common causes and fixes:
- Firewall/security group not allowing 18789: On cloud servers add an inbound rule for TCP 18789; on local Linux with ufw use
ufw allow 18789and reload. - Node version too old: If Node is below 22+, install or run may fail; switch with
nvm use 22or upgrade via the system package manager and retry. - Default token left unchanged and scanned: After deploy, change the default Gateway token in the Dashboard to avoid public scanning and abuse; also check system prompt length to avoid unusual token usage.
FAQ
Can I really get to first chat in 10 minutes?
With Node 22+ and normal network, using the install script or npm global install and following onboard → gateway → open 18789 console → send one command, most users can finish in about 10 minutes. For network or permission issues, see our OpenClaw troubleshooting and deploy articles.
Why deploy OpenClaw on a remote Mac?
A remote Mac gives you a full GUI (VNC), so system permission dialogs and Keychain prompts during onboard can be clicked directly, avoiding the case where SSH-only access cannot handle dialogs; this pairs well with our other OpenClaw deploy and permission-dialog articles.
What does openclaw doctor do?
openclaw doctor runs a health check and helps diagnose environment, dependencies, and config; if the console won’t start or behaves oddly, run doctor first and then use our troubleshooting guide.
Summary: You can get OpenClaw running on local Windows, Linux, or your own VPS, but Node version, port rules, permission dialogs, and environment issues often add troubleshooting time. If you want less environment hassle and a quick way to use OpenClaw (including GUI onboard and system permissions), rent a remote Mac from VNCMac and deploy OpenClaw in the VNC desktop to avoid local differences and unclickable dialogs; you get a ready Mac and a graphical setup in minutes. For options see our “OpenClaw environment choice” and “Quick start” pages.