How to Run OpenClaw on Mac: Laptop, Mac Mini, and 24/7 Server Setup

Hosting & Deployment · 12 min read

How to Run OpenClaw on Mac: Laptop, Mac Mini, and 24/7 Server Setup

Last updated: February 2026 · Reading time: 12 minutes · Difficulty: Beginner-friendly

Mac is OpenClaw's home turf. Peter Steinberger built it on a Mac, the community runs it on Macs, and macOS gets exclusive features that Linux and Windows don't — native iMessage integration, Apple Notes access, Reminders, and a menu bar companion app.

The Mac Mini M4 has become the unofficial reference hardware. At $599 it's a one-time purchase that eliminates monthly hosting costs, runs silent on under 10W at idle, and Apple Silicon's unified memory architecture handles local AI models better than most alternatives.

This guide covers three scenarios: running OpenClaw on your daily Mac (quick start, not 24/7), setting up a dedicated Mac Mini as a headless AI server, and the macOS-specific features and gotchas you need to know about.

Quick Start: OpenClaw on Your Daily Mac

If you just want to try OpenClaw on the Mac you already own, this takes about 10 minutes.

Step 1: Install Homebrew and Node.js

Open Terminal (Spotlight → type "Terminal") and run:

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

Follow the prompts. Then install Node.js:

brew install node@22

Verify with node --version — you need 22 or higher.

Step 2: Install OpenClaw

npm install -g openclaw@latest

Step 3: Run the Onboarding Wizard

openclaw onboard

The wizard walks you through everything — model selection, API keys, messaging channels. For a first-time setup, choose QuickStart and Claude Sonnet. You'll be chatting with your assistant in minutes.

Step 4: Grant macOS Permissions

This is the part that trips up most Mac users. OpenClaw needs three permissions to work properly:

  1. Full Disk Access — so it can read and write files
  2. Accessibility — so it can interact with apps
  3. Screen Recording — so it can see what's on screen

Go to System Settings → Privacy & Security and add Terminal to each of those three categories.

Critical: After toggling these on, quit and reopen Terminal. The permissions don't take effect until you restart the app. This catches almost everyone.

On macOS Sequoia (15+), Screen Recording may prompt you again each time you open Terminal. It's annoying, but it's Apple's security model.

Step 5: Start the Gateway

openclaw

Your assistant is running. Send a message through WhatsApp, Telegram, or whatever channel you configured during onboarding.

The limitation: When you close Terminal or shut down your Mac, the assistant stops. For a laptop you use daily, this is fine — OpenClaw runs when you're working and sleeps when you're not. For 24/7 operation, you need a dedicated setup.

Dedicated Mac Mini: The 24/7 AI Server

The Mac Mini M4 is the community's preferred hardware for always-on OpenClaw. Here's how to set it up properly.

Which Mac Mini to Buy

Base M4 (16GB RAM, 256GB SSD) — $599. This is enough for most people. OpenClaw itself uses 2–3GB of RAM. The rest is headroom for the browser tool, skills, and background processes. If you're only using cloud models (Claude, GPT, etc.), 16GB is plenty.

M4 with 32GB RAM — $799. Get this if you plan to run local AI models through Ollama alongside OpenClaw. A 7B parameter model needs roughly 8GB, and you want headroom for everything else.

Storage: The base 256GB is tight if you plan to keep multiple local models downloaded (they range from 4–40GB each). 512GB or more is recommended for Ollama users. OpenClaw itself uses under 1GB.

Essential Accessories

HDMI dummy plug ($8–12). This is critical. It tricks macOS into thinking a monitor is connected, which ensures stable operation when running headless (without a display). Without it, macOS can have resolution issues and Screen Sharing may not work properly. Buy one.

Ethernet cable or USB-C hub with Ethernet. Wi-Fi works, but Ethernet is more reliable for a device that needs to stay connected 24/7.

Keyboard with Touch ID (for initial setup only). Makes the first boot and configuration faster. After setup, you manage everything remotely.

Initial macOS Configuration

Then configure these settings:

Enable FileVault disk encryption. System Settings → Privacy & Security → FileVault. This encrypts your entire drive. If someone physically steals your Mac Mini, they can't read your data.

Enable the macOS firewall. System Settings → Network → Firewall. Turn it on.

Enable Remote Login (SSH). System Settings → General → Sharing → Remote Login. This is how you'll manage the Mac Mini from another computer.

Enable Screen Sharing. System Settings → General → Sharing → Screen Sharing. Useful for occasional graphical tasks.

Prevent sleep. System Settings → Battery → Options → enable "Wake for network access." Also: System Settings → Displays → Advanced → disable "Automatically adjust brightness." For full sleep prevention, install Amphetamine from the Mac App Store (free) and set it to keep the Mac awake indefinitely.

Set automatic login. System Settings → Users & Groups → Login Options → set your user to log in automatically. After a power outage and reboot, you want the Mac to come back up without waiting for a password.

Install OpenClaw

SSH into the Mac Mini from another machine:

ssh username@mac-mini.local

Or use the IP address if Bonjour isn't finding it. Then follow the same installation steps as above:

# Install Homebrew
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

# Install Node.js
brew install node@22

# Install OpenClaw
npm install -g openclaw@latest

# Run onboarding
openclaw onboard

Grant the three macOS permissions through Screen Sharing (VNC) since you need to click through System Settings.

Auto-Start on Boot

Create a LaunchAgent so OpenClaw starts automatically when the Mac boots and restarts if it crashes:

mkdir -p ~/Library/LaunchAgents

Create the plist file:

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN"
 "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
  <key>Label</key>
  <string>ai.openclaw.gateway</string>
  <key>ProgramArguments</key>
  <array>
       <string>/usr/local/bin/openclaw</string>
  </array>
  <key>RunAtLoad</key>
  <true/>
  <key>KeepAlive</key>
  <true/>
  <key>StandardOutPath</key>
  <string>/tmp/openclaw.stdout.log</string>
  <key>StandardErrorPath</key>
  <string>/tmp/openclaw.stderr.log</string>
  <key>WorkingDirectory</key>
  <string>/Users/YOUR_USERNAME</string>
</dict>
</plist>

Replace YOUR_USERNAME with your actual macOS username. Adjust the path to the openclaw binary if needed (check with which openclaw).

Load it:

launchctl load ~/Library/LaunchAgents/ai.openclaw.gateway.plist

OpenClaw now starts on boot and restarts automatically if it crashes. Verify with:

launchctl list | grep openclaw

Remote Access with Tailscale

You need to access the OpenClaw dashboard from outside your home network. Don't expose port 18789 to the internet — that's the gateway port and it should never be publicly accessible.

Instead, use Tailscale. It creates a secure VPN between your devices with zero port forwarding:

  1. Install Tailscale on the Mac Mini: brew install tailscale
  2. Install Tailscale on your phone/laptop
  3. Both devices are now on the same virtual network

Access the OpenClaw dashboard at http://mac-mini-tailscale-ip:18789 from anywhere, securely.

Tailscale is free for personal use (up to 100 devices). It's the community-recommended way to access OpenClaw remotely.

macOS-Exclusive Features

These features only work on macOS and are a big part of why Mac is the preferred platform.

iMessage Integration

OpenClaw can send and receive iMessages — the only self-hosted AI assistant that does this. If your contacts use iPhones, you can chat with your assistant through the Messages app with blue bubbles.

This requires the Mac to be signed into iCloud with iMessage enabled. The integration uses macOS's native Messages framework. No third-party libraries, no ToS violations.

Apple Notes and Reminders

Your assistant can read and write Apple Notes and create Reminders. "Add a note about today's meeting" or "remind me to call the dentist on Thursday" works natively through macOS APIs.

Menu Bar Companion App

OpenClaw has a macOS menu bar app that shows gateway status, active sessions, and quick controls. It's more convenient than checking the terminal or web dashboard for quick status checks.

Peekaboo (Screen Capture)

The Peekaboo skill captures and interprets what's on your screen. "What's on my screen right now?" or "read the error message in that dialog box." Requires the Screen Recording permission to be granted.

Apple Shortcuts Integration

You can trigger OpenClaw actions through Apple Shortcuts, which means Siri can relay commands to your assistant. "Hey Siri, tell my assistant to check my emails" is a real workflow some people use.

Mac Mini vs VPS: Which Should You Choose?

Both work well for 24/7 operation. Here's the trade-off.

Mac Mini Cloud VPS
Upfront cost $599 (one-time) $0
Monthly cost ~$5 electricity $6–12/month
Break-even ~8 months vs $12/mo VPS N/A
Privacy Data stays in your home Data on provider's servers
iMessage Yes No
Apple Notes/Reminders Yes No
Local AI models Yes (Ollama) Yes (but needs more RAM)
Maintenance You handle hardware Provider handles hardware
Uptime Depends on your power/internet 99.9%+ SLA
Portability Fixed to your home Access from anywhere natively

Choose Mac Mini if: You want macOS-exclusive features (iMessage, Notes, Reminders), you value data privacy, you plan to run local models, or you prefer a one-time cost over monthly bills.

Choose a VPS if: You want guaranteed uptime, you don't need macOS features, you don't want to manage hardware, or you're not ready to spend $599 upfront.

Many people start with a VPS to test OpenClaw, then buy a Mac Mini once they're committed. Our hosting comparison covers the best VPS options with pricing and free credit offers.

Troubleshooting macOS Issues

"Permission denied" errors

You didn't restart Terminal after granting permissions. Quit Terminal completely (Cmd+Q, not just close the window) and reopen it. This is the single most common issue on Mac.

Gateway won't start after macOS update

Major macOS updates occasionally break Node.js or npm global packages. Reinstall:

brew reinstall node@22
npm install -g openclaw@latest

High memory usage

Check if the browser tool is running with multiple tabs open. Each browser session consumes memory. Close unused sessions with the dashboard or restart the gateway.

If you're running Ollama alongside OpenClaw, local models are the likely culprit. A 7B model uses ~8GB. Monitor with top or Activity Monitor via Screen Sharing.

WhatsApp disconnects overnight

macOS's power management can interfere with network connections during sleep. Make sure "Wake for network access" is enabled in Battery settings. Amphetamine (free from Mac App Store) is the reliable solution — set it to prevent sleep entirely.

Screen Sharing shows low resolution

Plug in an HDMI dummy plug. Without it, macOS defaults to a low resolution when no display is connected. A $10 dummy plug solves this permanently.

OpenClaw doesn't start on boot

Check your LaunchAgent:

launchctl list | grep openclaw

If it's not running, check the error log:

cat /tmp/openclaw.stderr.log

Common causes: wrong path to the openclaw binary, wrong username in the plist, or the LaunchAgent not being loaded. Reload with:

launchctl unload ~/Library/LaunchAgents/ai.openclaw.gateway.plist
launchctl load ~/Library/LaunchAgents/ai.openclaw.gateway.plist

Running Local Models with Ollama

One of the Mac Mini's biggest advantages is running AI models locally through Ollama — zero API costs, complete privacy.

# Install Ollama
brew install ollama

# Pull a model
ollama pull hermes-2-pro       # Good for tool calling
ollama pull llama3.3         # Strong general model
ollama pull mistral         # Fast, lightweight

{
    "agent": {
        "model": "ollama/hermes-2-pro"
    }
}

Apple Silicon's unified memory means the GPU and CPU share the same RAM pool, which is a genuine advantage for local inference. A Mac Mini with 32GB RAM can comfortably run a 7B model while OpenClaw handles everything else.

Be realistic about quality. Local 7B models don't match Claude Sonnet for agentic tasks. Tool calling is less reliable, reasoning is shallower, and context windows are much shorter. Local models work for simple tasks and experimentation. For a production assistant, cloud models are significantly better.

A practical setup: use Claude Sonnet as your primary model and Ollama as a fallback for when you want to tinker without API costs. See our model comparison guide for the full breakdown.

Frequently Asked Questions

Can I run OpenClaw on a MacBook full-time?

Technically yes, but it's not ideal. Laptops sleep when you close the lid, pause when the battery gets low, and you're tying up your daily machine. A dedicated Mac Mini is the better long-term solution. A MacBook works fine for testing and occasional use.

Do I need the M4 specifically?

No. OpenClaw runs on any Apple Silicon Mac (M1 and later) and even Intel Macs. The M4 Mac Mini is recommended because it's the cheapest current option with the best performance-per-watt, not because older Macs can't handle it.

Is 16GB RAM really enough?

For OpenClaw with cloud models, yes. The gateway uses 2–3GB, plus whatever the browser tool and skills consume. You'll rarely exceed 8GB total. Local models through Ollama are the reason to upgrade to 32GB.

Can I run OpenClaw in Docker on Mac?

Yes. Docker Desktop and OrbStack both work on macOS. This adds an extra layer of isolation — OpenClaw runs in a container with only the access you explicitly mount. See our Docker guide for the containerized setup.

How loud is the Mac Mini when running OpenClaw 24/7?

Silent for cloud model use. The fan almost never spins up because OpenClaw is mostly waiting for API responses. If you're running local models through Ollama, you'll hear the fan during inference. Under normal assistant use, it's quieter than a refrigerator.

Just getting started? Follow our setup guide for a platform-agnostic walkthrough. Prefer cloud hosting? Check our VPS comparison for the best deals. Want to understand the full costs? See our complete breakdown.