2026 OpenClaw from install to first chat: 10-minute route map, port 18789 and common pitfalls

2026 OpenClaw from Install to First Chat: 10-Minute Route Map (Port 18789 & Common Pitfalls)

~12 min read
OpenClaw 10 min guide openclaw onboard Port 18789

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 -v in 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.
Tip 1: Most “cannot reach http://IP:18789” issues are due to the firewall not allowing 18789; on cloud servers add an inbound rule in the security group.
Tip 2: If npm install hangs or times out, try a mirror: npm config set registry https://registry.npmmirror.com.
Tip 3: When deploying OpenClaw on a remote Mac via VNC, you can complete onboard and permission dialogs in the GUI, avoiding the case where SSH-only access cannot click system dialogs; see our OpenClaw deploy and permission-dialog articles.

Steps from install to first chat

Using the install script and onboard, the minimal path is as follows.

1

Install OpenClaw

Run curl -fsSL https://openclaw.ai/install.sh | bash or npm install -g openclaw@latest and wait for completion.

2

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.

3

Start Gateway

Run openclaw gateway (or use systemd/pm2 for persistence); ensure port 18789 is listening.

4

Open web console

In a browser go to http://localhost:18789 (or your server IP:18789) to open the Dashboard.

5

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:

  1. Firewall/security group not allowing 18789: On cloud servers add an inbound rule for TCP 18789; on local Linux with ufw use ufw allow 18789 and reload.
  2. Node version too old: If Node is below 22+, install or run may fail; switch with nvm use 22 or upgrade via the system package manager and retry.
  3. 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.

Choose your Mac node and OpenClaw deployment

Deploy OpenClaw on a remote Mac via VNC for easy onboard and permission dialogs; get going in minutes with pay-as-you-go.

  • VNC desktop supports system dialogs and Keychain for first-time OpenClaw setup
  • Fits with our OpenClaw deploy and troubleshooting articles
  • Hourly or monthly billing; no need to own a Mac or maintain a local environment