Install OpenClaw — Free Step-by-Step Guide for Mac Mini & VPS

A free install OpenClaw service tutorial written for users in Malaysia. Pick the path that fits — a Mac mini at home or in the office, or a Linux VPS in the cloud — and follow the exact commands below to get your own self-hosted AI assistant running on WhatsApp, Telegram, or Slack.

Stuck on a step? WhatsApp us with the step number and any error output. We will advise for free — no obligation.

WhatsApp Us for Free Install Advice

Before You Start — Pick Your Path

OpenClaw runs the same way on every machine — the difference is where you host it. Pick Mac mini if you want a quiet home / office box you control. Pick VPS if you want cloud-hosted, accessible from anywhere, no physical machine.

Path A — Mac mini

Self-hosted at home or office in KL

Best for individuals and SMEs who already own a Mac mini, want data on-prem, and don't mind the box staying plugged in 24/7. Apple Silicon (M1 / M2 / M4) with 16 GB RAM is the sweet spot. Budget ~60 minutes.

Jump to Mac mini steps ↓
Path B — Linux VPS

Cloud-hosted, accessible anywhere

Best for teams, remote-first users, or anyone without spare hardware. Pick a Singapore region (DigitalOcean SGP1, Linode Singapore, AWS ap-southeast-1) for low MY latency. Minimum 2 vCPU / 4 GB RAM. Budget ~90 minutes.

Jump to VPS steps ↓

Path A — Install OpenClaw on Mac Mini

These steps follow the official OpenClaw macOS onboarding flow. You will need the Mac mini powered on, signed in to an admin account, and connected to your home / office Wi-Fi.

Authoritative reference: docs.openclaw.ai / macOS

1

Install the OpenClaw macOS app

Download from the official site and drag to /Applications. On first launch, macOS will prompt for TCC permissions — Notifications, Accessibility, Screen Recording, Microphone, Speech Recognition, and Automation/AppleScript. Grant all; the gateway uses them to read notifications and control apps on your behalf.

2

Install the CLI

Open Terminal and install OpenClaw globally via npm (install Node.js 20+ first if you don't have it, via brew install node):

npm install -g openclaw@latest

Verify with openclaw --version.

3

Run the onboarding wizard

This installs the launchd daemon so the gateway restarts on reboot, and walks you through picking a model provider (Claude, Claude Max, ChatGPT, DeepSeek, or local Ollama):

openclaw onboard --install-daemon

When the wizard asks for a bind mode, pick Local — the gateway stays on loopback (127.0.0.1). Do not expose it directly to the LAN.

4

Connect your messaging platform

For WhatsApp (recommended in MY — use a dedicated +60 number, not your primary WhatsApp):

openclaw channels login --channel whatsapp

A QR code will appear. Scan it from WhatsApp → Linked Devices. For Telegram substitute --channel telegram and follow the bot-token prompt.

5

Lock down who can DM your assistant

Edit ~/.openclaw/openclaw.json and set an allowlist so only your number(s) can talk to the gateway:

{
  "channels": {
    "whatsapp": {
      "dmPolicy": "allowlist",
      "allowFrom": ["+60123456789"],
      "groupPolicy": "disabled"
    }
  }
}

Restart the gateway: openclaw gateway restart.

6

Configure for 24/7 headless operation

System Settings → Energy → uncheck "Put hard disks to sleep when possible" and set "Prevent automatic sleeping when the display is off" to ON. Enable FileVault (System Settings → Privacy & Security) and turn on the macOS firewall. For remote access into the Mac from outside your LAN, install Tailscale and use tailscale serve — don't port-forward the gateway on your router.

7

Open the dashboard and test

From any machine on the same Tailscale network:

openclaw dashboard

Then WhatsApp "Hello" to the linked number. If the assistant replies, you're done.

Path B — Install OpenClaw on Linux VPS

Written against Ubuntu 22.04 / 24.04 on any provider (DigitalOcean, Linode, Hetzner, AWS Lightsail). For MY latency, a Singapore region is the right call.

Authoritative reference: docs.openclaw.ai / vps

1

Provision the VPS

Minimum: 2 vCPU, 4 GB RAM, 20 GB SSD. Region: Singapore (ap-southeast-1 on AWS, SGP1 on DO). SSH in as a non-root user with sudo, and make sure the box is fully updated:

sudo apt update && sudo apt upgrade -y
2

Install Node.js 20+

Using NodeSource's official script:

curl -fsSL https://deb.nodesource.com/setup_20.x | sudo -E bash -
sudo apt install -y nodejs

Confirm with node --version — should be v20.x or higher.

3

Install OpenClaw

npm install -g openclaw@latest

If you're on an ARM VPS or a small droplet, also add these to ~/.bashrc to speed up node restarts (per the official runbook):

export NODE_COMPILE_CACHE=/var/tmp/openclaw-compile-cache
mkdir -p /var/tmp/openclaw-compile-cache
export OPENCLAW_NO_RESPAWN=1
4

Run onboarding + install the systemd user service

openclaw onboard --install-daemon

The wizard installs openclaw-gateway.service as a user-scope systemd unit and enables auto-start on reboot.

5

Configure the gateway to loopback-only

The secure default is to keep the gateway on 127.0.0.1 and reach it via SSH tunnel or Tailscale — not to expose it on the public internet.

openclaw configure

Pick Local (this machine). If you bind to LAN or tailnet instead, OpenClaw will require you to set gateway.auth.token or gateway.auth.password — don't skip this.

6

Connect WhatsApp or Telegram

openclaw channels login --channel whatsapp

On a headless VPS the QR code renders in the terminal as ASCII — scan it from WhatsApp → Linked Devices on your phone.

7

Set the allowlist + restart

Same config as the Mac mini path — edit ~/.openclaw/openclaw.json and lock dmPolicy to allowlist, then:

systemctl --user restart openclaw-gateway.service
8

Access the dashboard over SSH tunnel

ssh -L 8787:127.0.0.1:8787 user@your-vps-ip

Then visit http://127.0.0.1:8787 in your local browser. WhatsApp "Hello" to the linked number — you're live.

Security Checklist — Don't Skip

An OpenClaw gateway can read files, send messages, and call tools on your behalf. A leaked gateway is a leaked assistant. Five minimum controls before you hand out your WhatsApp number:

1. Loopback-only gateway

Keep the gateway bound to 127.0.0.1. Remote access is done via SSH tunnel or Tailscale Serve — never by exposing the port publicly. If you must bind to LAN / tailnet, require gateway.auth.token.

2. Allowlist your DMs

Set dmPolicy: "allowlist" and list only the E.164 phone numbers you actually want to talk to the assistant from. Default open means anyone who finds your linked number can send prompts.

3. Disk encryption at rest

On Mac mini, turn on FileVault. On VPS, enable full-disk encryption if your provider offers it (Linode, Hetzner), and encrypt the ~/.openclaw/ state dir at minimum.

4. Firewall + SSH hardening

macOS firewall ON. On VPS, ufw default-deny inbound except port 22, disable password SSH (PasswordAuthentication no), use keys only. Install fail2ban.

5. Separate the WhatsApp number

Don't link your primary personal WhatsApp. Use a dedicated Malaysian +60 number (prepaid SIM or second eSIM). If the gateway is ever compromised, you can unlink without touching your main account.

6. Rotate model API keys quarterly

Your Claude / OpenAI / DeepSeek key is stored in OpenClaw's secrets file. Treat it like a password — rotate on a schedule, and monitor the provider's usage dashboard for anomalies.

Common Install Questions

Yes. The tutorial is free to follow, and OpenClaw itself is free and open source. If you get stuck on any step, WhatsApp us and we will advise at no cost. We only charge (RM 2,000, one-time) if you want us to do the whole install and configuration for you — see the done-for-you setup package.

Yes. OpenClaw supports Claude (including Claude Max subscribers), ChatGPT, DeepSeek, and local models via Ollama. During the openclaw onboard wizard, pick Claude as the model provider and paste the API key or OAuth credentials tied to your Max plan. Billing continues to come from Anthropic — OpenClaw is just the gateway that routes your WhatsApp / Telegram messages into Claude and back.

Yes, by pointing OpenClaw at the Memos REST API as a tool endpoint. Memos exposes an HTTP API for creating and searching notes, so you can configure an OpenClaw workflow that writes every message you forward to your assistant into a Memos entry, or pulls matching notes back on request. This is a custom workflow step — not bundled by default, and a common reason people WhatsApp us when their DIY "memos openclaw" setup gets stuck on the auth token.

Three to watch. (1) Latency: Claude, OpenAI, and most frontier-model endpoints are US or EU hosted — expect 180–250 ms extra round-trip from MY. For a Mac mini at home in KL this is fine; for a VPS, pick Singapore (DigitalOcean SGP1, Linode Singapore, AWS ap-southeast-1). (2) WhatsApp number: Use a Malaysian +60 number that is not your primary personal line — WhatsApp may flag a new linked device on your main account. (3) Language: Configure BM / Mandarin / mixed-language system prompts during onboarding — the default prompts assume English only.

Check the logs first: journalctl --user -u openclaw-gateway.service -n 200 --no-pager. Ninety percent of VPS failures are one of: Node.js below v20 (upgrade), port 8787 already in use (change the port in openclaw.json), or ~/.openclaw/ on a cloud-synced drive (move it local, set OPENCLAW_STATE_DIR=~/.openclaw). If the logs mean nothing to you, WhatsApp them to us — we'll translate.

Singapore-region VPS with 2 vCPU / 4 GB RAM runs USD 12–24 / month (roughly RM 55–110). Add the model cost: Claude and ChatGPT pay-as-you-go typically lands at RM 30–80 / month for light SME use; Claude Max flat rate is higher but uncapped. If you want zero ongoing model cost, run Ollama with Llama 3 locally — RAM-heavy but free after install.

Yes — BixTech offers a done-for-you OpenClaw install service for RM 2,000 one-time. It covers the discovery call, full install on your Mac mini or VPS, messaging platform integration, 3–5 custom workflows for your role, a one-hour training session, and 14 days of WhatsApp support after handover. See the full OpenClaw setup package.

Hit a Wall? Two Ways Forward.

Both paths start on WhatsApp. The free-advice one stays free. The done-for-you one takes RM 2,000 off your weekend.

Free — Install advice

Tell us the step number and paste any error output. We respond with the fix. No sales pitch, no obligation.

WhatsApp for Free Advice

RM 2,000 — Done-for-you install

Skip the weekend. We install on your Mac mini or VPS, configure WhatsApp / Telegram, build 3–5 workflows, and train your team. 14-day support included.

See the Setup Package →

Or email [email protected] · Call +60 11-1300 4052