Purpose: I wish to show that I can repurpose an old computer for a new purpose. This project is to show I can run OpenClaw (the trending autonomous AI agent, formerly known as Clawdbot or Moltbot) on my Intel Mac Mini. This is my plan for this project and I would appreciate any feedback. I will start the project this week and hope to complete it and have OpenClaw running by the weekend.

Part 1 Installing OpenClaw

Introduction:

To run OpenClaw (the trending autonomous AI agent, formerly known as Clawdbot or Moltbot) on your Intel Mac Mini, you’ll need to set up a small local environment. Since you are on an Intel Mac, the process is very stable, but ensure you are running at least macOS 13 (Ventura) or newer for the best results.

Problem 1: Mac OS likes to go to sleep to save energy

Solution: To run OpenClaw (the trending autonomous AI agent, formerly known as Clawdbot or Moltbot) on your Intel Mac Mini, you’ll need to set up a small local environment. Since you are on an Intel Mac, the process is very stable, but ensure you are running at least macOS 13 (Ventura) or newer for the best results.

Step 1: Prevent Your Mac from Sleeping

For an AI agent to stay "alive" and respond to messages (like Telegram or WhatsApp) 24/7, your Mac Mini cannot go to sleep.

• The Recommended App: Amphetamine (Free on the Mac App Store).
• It is the gold standard for "keep-awake" utilities.
• Configuration: Once installed, click the pill icon in your menu bar and select "Indefinite." * Pro Tip: In Amphetamine's settings, enable "Allow display sleep while computer is awake" to save your monitor's life while the Mac Mini keeps processing in the background.

• The "Built-in" Way: Open Terminal and type caffeinate. As long as that Terminal window is open, your Mac will not sleep. However, Amphetamine is much more reliable for a 24/7 setup.

Step 2: Install Prerequisites

Open your Terminal (Cmd + Space, type "Terminal") and run these commands:

1. Install Homebrew (The Mac package manager):

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

2. **Install Node.js 22+**:
OpenClaw requires a modern version of Node.

```bash
brew install node

Verify with node -v. It should be v22.x.x or higher.

---

Step 3: Download and Install OpenClaw

The OpenClaw team provides a one-line installer that handles the heavy lifting.

1. Run the Installer:

curl -fsSL https://openclaw.ai/install.sh | bash

1. Run the Onboarding Wizard:
   This is where you'll connect your "brain" (API keys from OpenAI, Anthropic, or Gemini) and your "ears" (Telegram, WhatsApp, etc.).

openclaw onboard --install-daemon

• The --install-daemon flag is crucial for your Mac Mini. It creates a background service (launchd) so OpenClaw starts automatically whenever you turn the computer on.

---

Step 4: Connect Your Channels

After onboarding, you'll need to "pair" your messaging apps so you can talk to your agent from your phone.

• For Telegram: Create a bot via @BotFather, get the token, and follow the prompts in the terminal.
• For WhatsApp: The terminal will generate a QR Code. Scan it with your WhatsApp app (Linked Devices) just like you would for WhatsApp Web.

---

Step 5: Verify the Setup

Check if the "Gateway" (the bridge between your Mac and the AI) is running:

openclaw status

If you see "Active," you're good to go! You can now message your bot from your phone, and your Intel Mac Mini will do all the thinking.

Summary of "The 2026 AI Agent Stack"

Component	Tool
Hardware	Intel Mac Mini (Stable & Reliable)
Keep-Awake	Amphetamine (Indefinite Session)
Runtime	Node.js 22
Agent Core	OpenClaw (Local Gateway)
Remote Access	Telegram / WhatsApp

Here is a related vidoe I found searching YouTube

Running OpenClaw locally on your Mac
This video provides a visual walkthrough of the installation steps and how to manage the background services on macOS.

Part 2 installing a local large language model AI to reduce cost on online LLM access fees

Introduction:

Connecting Ollama to OpenClaw on your Mac Mini is a great move—it gives you a 100% private AI assistant with zero subscription fees. Because OpenClaw is designed to be "OpenAI-compatible" but not "OpenAI-dependent," you can easily swap the expensive cloud brain for a local one.

Here is the step-by-step guide to making them work together.

---

Step 1: Install and Prepare Ollama

If you haven't already, install Ollama and download a model that is strong enough to handle "Agentic" tasks (like following multi-step instructions).

1. Download Ollama: Get it at ollama.com.
2. Pull a "Tool-Capable" Model: For an Intel Mac Mini, I recommend Qwen 2.5 (7B or 14B) or Llama 3.3. These models are excellent at "Function Calling," which OpenClaw needs to actually control your computer.

• Open Terminal and run:

ollama pull qwen2.5-coder:7b

1. Ensure it's Running: Make sure the Ollama icon is in your Mac's menu bar.

---

Step 2: The "One-Command" Connection

In early 2026, the fastest way to link them is using the new launch command which automatically handles the configuration for you.

1. Run the Launch Command:

ollama launch openclaw --model qwen2.5-coder:7b

• What this does: Ollama will detect your OpenClaw installation, set itself as the "Provider," and tell OpenClaw to use the local API endpoint (http://127.0.0.1:11434) instead of OpenAI.

---

Step 3: Manual Configuration (If needed)

If the auto-launcher doesn't pick up your models, you can manually set the "Ollama API Key" environment variable. Even though Ollama doesn't actually require a key, OpenClaw looks for one to verify the connection is active.

1. Set the Key:

openclaw config set models.providers.ollama.apiKey "ollama-local"

1. Verify the Link:

openclaw models list

• You should now see your local Ollama models (e.g., ollama/qwen2.5-coder:7b) appearing in the list alongside any cloud models.

---

Step 4: Set Your Local Model as the "Primary"

To make sure OpenClaw always uses your Mac Mini's power first:

1. Open the OpenClaw Control Center in your browser (usually http://localhost:18789).
2. Go to Settings > Agents.
3. Change the Primary Model to your local Ollama model.

---

Important Tips for Intel Mac Mini Users I found online

• Context Window: OpenClaw sends a lot of data (the "System Prompt") to the AI. If your bot starts acting "forgetful," increase the context window in your config to at least 32,000 (32k) or 64k if your RAM allows.
• Memory Usage: If you have an 8GB RAM Mac Mini, stick to 7B or 8B models. If you have 16GB or more, try a 14B model for significantly better reasoning.
• The "Silent Failure" Fix: If the bot doesn't respond, ensure your base URL in the config is http://127.0.0.1:11434 (no /v1 at the end).

Setup OpenClaw with Ollama for a zero-cost AI assistant

This video is a complete walkthrough for developers and enthusiasts who want to run a secure, local-only AI workflow using OpenClaw and Ollama.

Part 3 Messaging Component

Introduction:

The third component you need is a Messaging Bridge (also called a Connector or Channel).
While OpenClaw is the body and Ollama is the brain, you still need a way to talk to the agent when you aren't sitting at your Mac Mini. Without a bridge, your AI is essentially a genius trapped in a room with no door.

The Missing Link: Your Messaging Bot

Because OpenClaw is designed to be a "headless" agent (meaning it runs in the background), you interact with it through everyday apps. Most users choose Telegram or WhatsApp.

Why Telegram is the Best Choice for Your Setup:

• It’s Free: No monthly fees for bot accounts.
• It’s Persistent: Since your Mac Mini is always on (thanks to Amphetamine), you can text your bot from your iPhone, Android, or even a web browser while you're at work, and it will execute tasks on your Mac Mini at home.
• File Support: You can send a PDF denial template to your Telegram bot, and the agent on your Mac Mini will "see" it, process it using Ollama, and send the finished Word document back to you.

---

How to Set Up the Third Component

1. Create the Bot: Open Telegram and search for @BotFather.
2. Get the Token: Type /newbot and follow the instructions to name it (e.g., "MyLocalClawBot"). It will give you an API Token (a long string of numbers and letters).
3. Link to OpenClaw: In your Mac Terminal, run:

openclaw onboard --channel telegram

1. Paste the Token: When prompted, paste the token from @BotFather.

---

Your Final "Local AI Stack"

Component	What it is	Role
1. OpenClaw	The "Body"	The software that actually moves files and runs scripts on your Mac.
2. Ollama	The "Brain"	The local engine that processes your logic and keeps data private.
3. Telegram Bot	The "Mouth/Ears"	The remote control that lets you talk to your Mac from anywhere.

---

Connecting OpenClaw to Telegram

This video walk-through shows exactly how to use BotFather to get your token and how to enter the pairing code to finish the connection.

One Last (Optional) Tip: Tailscale

If you want to access the OpenClaw Dashboard (the visual interface) from outside your home without opening dangerous ports on your router, many users install Tailscale. It’s a "Zero-Config VPN" that lets you securely log into your Mac Mini's local dashboard as if you were sitting right in front of it.

More on Tailscale

Introduction:

Tailscale is the "secret sauce" for home server enthusiasts. In simple terms, it creates a private, encrypted tunnel between your devices (your Mac Mini, your iPhone, and your laptop) so they act like they are on the same home Wi-Fi, even if you are halfway across the world.

For your OpenClaw setup, it solves a major problem: How do you access the dashboard on your Mac Mini without opening a hole in your home router?

---

1. Why Tailscale is Better than "Port Forwarding"

Normally, to access a home computer from the internet, you have to "open a port" on your router. This is like leaving a window unlocked in your house—eventually, a hacker's bot will find it and try to get in.

• Tailscale is "Zero-Exposure": It doesn't open any ports. It uses a technology called WireGuard to create a secure, direct connection between your phone and your Mac. To the rest of the internet, your Mac Mini remains completely invisible.

2. How to Set It Up for OpenClaw

It takes about 3 minutes to set up:

1. Install on Mac Mini: Download the Standalone variant from tailscale.com (it's better than the App Store version for servers). Log in.
2. Install on your Phone/Laptop: Get the Tailscale app on your iPhone or Android and log into the same account.
3. Get your "Magic" IP: Open the Tailscale app on your Mac. You’ll see an IP address starting with 100.x.x.x. This is your Mac's private address that stays the same forever.
4. Access OpenClaw: While away from home, turn on Tailscale on your phone. Open your browser and type in that 100.x.x.x address followed by the OpenClaw port (usually :18789).

• Example: http://100.12.34.56:18789

3. The "Funnel" Feature (Advanced)

As of 2026, OpenClaw actually has a built-in "Tailscale Mode." If you want to share your dashboard with a specific coworker or friend without them needing Tailscale, you can use Tailscale Funnel.

• This provides a "Public URL" (like https://macmini.tailnet-name.ts.net) that is protected by a password you set in OpenClaw. It’s the easiest way to get a "real" website address for your local AI.

---

Summary of the Benefit

Without Tailscale	With Tailscale
Can only use OpenClaw while at home.	Can use OpenClaw from a coffee shop, airport, or office.
Risks hackers finding your open ports.	100% invisible to anyone outside your private "Tailnet."
Must deal with complicated router settings.	Works instantly, even behind strict office firewalls.

How to install the Standalone version of Tailscale (not the App Store version) because it allows Tailscale to run even if you aren't logged into a user account.

Introduction:

As of the early 2026 update to OpenClaw (v2.x), there is a built-in "Tailscale Mode" that automates the networking for you. This allows OpenClaw to talk directly to your Tailscale app to handle the secure connection without you needing to touch your router.

Here are the terminal commands to set it up on your Mac Mini.

Step 1: Choose Your Mode

There are two ways to run this. Decide which one fits your needs:

• "Serve" Mode (Recommended): This makes the dashboard available only to your private devices (your phone/laptop). It is 100% private.
• "Funnel" Mode: This creates a public internet link (like https://your-mac.ts.net). This is great if you want to access it from a computer that doesn't have Tailscale installed (like a locked-down work PC).

---

Step 2: Run the Command

Open your Terminal and enter the following command to tell OpenClaw to use Tailscale:

For private "Serve" mode:

openclaw config set gateway.tailscale.mode "serve"

For public "Funnel" mode:
(Note: This requires you to have a password set on your OpenClaw gateway for security.)

openclaw config set gateway.tailscale.mode "funnel"

---

Step 3: Restart and Verify

For the changes to take effect, you must restart the OpenClaw service:

openclaw gateway restart

Now, run the "Status" command to see your new secure URL:

openclaw status --deep

Look for a line that says "Tailscale URL." It will look something like https://macmini.tail1234.ts.net. You can now paste that link into any browser on your phone, and it will take you straight to your OpenClaw dashboard!

---

Why this is a "Game Changer"

• SSL Included: Tailscale automatically provides an HTTPS certificate. This means your password and data are encrypted while traveling over the internet, and you won't get those "This site is not secure" warnings in your browser.
• Identity Headers: In "Serve" mode, OpenClaw can actually see who is logged into Tailscale. If it recognizes your Tailscale account, it can let you in automatically without asking for a token every time.

Important Note for Intel Macs: If you get an error saying "Tailscale not found," make sure you installed the Standalone version of Tailscale from their website. The App Store version is "sandboxed" and sometimes prevents OpenClaw from talking to it.

Creating a Watchdog Script

Introduction:

A "Watchdog" script is a simple background program that acts like a heartbeat monitor. Since OpenClaw v2.x (2026) has its own health-check API, we can write a script that checks two things:

1. Is the OpenClaw Gateway alive?
2. Is Tailscale connected?

If either one fails, the script will try to fix it and send you a Telegram message so you aren't left wondering why your bot isn't responding.

---

Step 1: Create the Watchdog Script

Open your Terminal and create a new file:

nano ~/openclaw_watchdog.sh

Paste this code into the editor (replace the placeholders with your actual info):

#!/bin/bash

# --- CONFIGURATION ---
# Your Telegram Bot Token from @BotFather
TOKEN="YOUR_TELEGRAM_BOT_TOKEN"
# Your Telegram User ID (Get it by messaging @userinfobot)
USER_ID="YOUR_TELEGRAM_USER_ID"
# Your Tailscale IP (The 100.x.x.x address)
TS_IP="YOUR_TAILSCALE_IP"

# --- THE CHECKS ---

# 1. Check Tailscale Connection
if ! /Applications/Tailscale.app/Contents/MacOS/Tailscale status | grep -q "100."; then
    MESSAGE="⚠️ ALERT: Tailscale is DOWN on the Mac Mini. Attempting to reconnect..."
    curl -s -X POST "https://api.telegram.org/bot$TOKEN/sendMessage" -d "chat_id=$USER_ID&text=$MESSAGE"
    # Try to restart Tailscale (Standalone version)
    sudo /Applications/Tailscale.app/Contents/MacOS/Tailscale up
fi

# 2. Check OpenClaw Gateway Health
# Using the 2026 health-check command
if ! openclaw health --json | grep -q '"status":"ok"'; then
    MESSAGE="❌ ALERT: OpenClaw Gateway has CRASHED. Restarting service now..."
    curl -s -X POST "https://api.telegram.org/bot$TOKEN/sendMessage" -d "chat_id=$USER_ID&text=$MESSAGE"
    openclaw gateway restart
fi

Press Ctrl+O, Enter, then Ctrl+X to save and exit.

---

Step 2: Make it Executable

You need to give the script permission to run:

chmod +x ~/openclaw_watchdog.sh

---

Step 3: Schedule it to Run Every 5 Minutes

On macOS, the easiest way to keep this running in the background is using crontab.

1. Open your cron settings:

crontab -e

1. Add this line at the very bottom:

*/5 * * * * /Users/YOUR_USERNAME/openclaw_watchdog.sh

(Replace YOUR_USERNAME with your actual Mac username. If you aren't sure, type whoami in the terminal.)

---

How This Protects You

• Instant Recovery: If your Mac Mini has a momentary Wi-Fi glitch and Tailscale drops, the script will force it back "up" before you even notice.
• Notification: You’ll get a Telegram message like: "❌ ALERT: OpenClaw Gateway has CRASHED. Restarting..." This lets you know that the "brain" was rebooted and is ready for your next command.
• Zero Maintenance: Because it’s a cron job, it starts automatically every time you turn on your Mac Mini.

Tip I found online for 2026

If you want to be extra secure, you can add a third check for Ollama. Just add these lines to the script:

if ! pgrep -x "ollama" > /dev/null; then
    curl -s -X POST "https://api.telegram.org/bot$TOKEN/sendMessage" -d "chat_id=$USER_ID&text=🧠 Ollama was closed. Reopening..."
    open ollama://run
fi