Updated Feb 10, 2026

Always On Duty

Your AI Employee works brilliantly when your computer is on and you are awake. But what happens at 3am when a client sends an urgent email? What about when you are traveling and your laptop is closed? Right now, your employee takes the day off whenever you do.

This lesson transforms your AI Employee from a work-hours assistant into a true 24/7 team member. You will learn two approaches: running persistently on your local machine, and deploying to cloud infrastructure that never sleeps. By the end, your AI Employee will be available every moment of every day, processing requests whether you are asleep, traveling, or simply away from your desk.

This is an advanced, optional lesson. If you are satisfied with your AI Employee working only when your computer is on, you can skip ahead to the Chapter Assessment. But if you want the full autonomous experience, the investment here pays dividends in availability.

The Always-On Promise

Human employees have limits. They sleep, take vacations, get sick. Your AI Employee has none of these constraints. The only limitation is infrastructure: it needs somewhere to run.

Your Current Setup	What Changes
Works when laptop is open	Works 24/7/365
Stops when you reboot	Survives reboots
Unavailable while traveling	Available everywhere
Dependent on your presence	Truly autonomous

There are two paths to always-on operation:

Option A: Local Always-On (PM2)

Your computer runs continuously
Process manager keeps gateway alive
Cost: $0 (electricity only)
Limitation: Computer must stay on

Option B: Cloud Always-On (Oracle Free)

Remote server runs your gateway
Available even when your computer is off
Cost: $0 (Oracle Always Free tier)
Limitation: Setup complexity, ARM architecture

Choose based on your needs. Many users start with Option A and graduate to Option B when they want true independence from their local machine.

Option A: Local Always-On with PM2

PM2 is a production process manager for Node.js applications. It keeps your gateway running, restarts it if it crashes, and can survive system reboots.

Installing PM2

PM2 installs globally via npm:

npm install -g pm2

Output:

added 182 packages in 8s

18 packages are looking for funding
  run `npm fund` for details

Verify the installation:

pm2 --version

Output:

5.3.0

Creating an Ecosystem File

PM2 uses an ecosystem file to define how your application runs. Create this in your home directory or a project folder:

nano ~/openclaw-ecosystem.config.js

Enter the following configuration:

module.exports = {
  apps: [{
    name: "openclaw-gateway",
    script: "openclaw",
    args: "gateway run --port 18789",
    watch: false,
    autorestart: true,
    max_restarts: 10,
    restart_delay: 5000,
    env: {
      NODE_ENV: "production"
    }
  }]
};

Save and exit.

What each setting does:

Setting	Purpose
`name`	Identifier for this process in PM2
`script`	The command to run (openclaw binary)
`args`	Arguments passed to the command
`watch`	Restart on file changes (disabled for stability)
`autorestart`	Restart if process crashes
`max_restarts`	Give up after 10 consecutive failures
`restart_delay`	Wait 5 seconds between restart attempts

Starting with PM2

Launch your gateway through PM2:

pm2 start ~/openclaw-ecosystem.config.js

Output:

[PM2] Spawning PM2 daemon with pm2_home=/Users/user/.pm2
[PM2] PM2 Successfully daemonized
[PM2] Starting /usr/local/bin/openclaw in fork_mode (1 instance)
[PM2] Done.
┌─────┬────────────────────┬─────────────┬─────────┬─────────┬──────────┬────────┬──────┬───────────┬──────────┬──────────┬──────────┬──────────┐
│ id  │ name               │ namespace   │ version │ mode    │ pid      │ uptime │ ↺    │ status    │ cpu      │ mem      │ user     │ watching │
├─────┼────────────────────┼─────────────┼─────────┼─────────┼──────────┼────────┼──────┼───────────┼──────────┼──────────┼──────────┼──────────┤
│ 0   │ openclaw-gateway   │ default     │ N/A     │ fork    │ 12345    │ 0s     │ 0    │ online    │ 0%       │ 45.2mb   │ user     │ disabled │
└─────┴────────────────────┴─────────────┴─────────┴─────────┴──────────┴────────┴──────┴───────────┴──────────┴──────────┴──────────┴──────────┘

Persisting Across Reboots

By default, PM2 processes disappear when your computer restarts. To make them survive reboots:

pm2 save
pm2 startup

Output:

[PM2] Saving current process list...
[PM2] Successfully saved in /Users/user/.pm2/dump.pm2
[PM2] To setup the Startup Script, copy/paste the following command:
sudo env PATH=$PATH:/usr/local/bin pm2 startup launchd -u user --hp /Users/user

Run the command PM2 provides (it will be specific to your system):

sudo env PATH=$PATH:/usr/local/bin pm2 startup launchd -u user --hp /Users/user

Now your gateway will start automatically when your computer boots.

Monitoring PM2 Processes

PM2 provides several monitoring commands:

Check status:

pm2 status

Output:

┌─────┬────────────────────┬─────────────┬─────────┬─────────┬──────────┬────────┬──────┬───────────┬──────────┬──────────┐
│ id  │ name               │ namespace   │ version │ mode    │ pid      │ uptime │ ↺    │ status    │ cpu      │ mem      │
├─────┼────────────────────┼─────────────┼─────────┼─────────┼──────────┼────────┼──────┼───────────┼──────────┼──────────┤
│ 0   │ openclaw-gateway   │ default     │ N/A     │ fork    │ 12345    │ 45m    │ 0    │ online    │ 0.1%     │ 62.3mb   │
└─────┴────────────────────┴─────────────┴─────────┴─────────┴──────────┴────────┴──────┴───────────┴──────────┴──────────┘

View logs:

pm2 logs openclaw-gateway

Output:

[TAILING] Tailing last 15 lines for [openclaw-gateway] process (change the value with --lines option)
/Users/user/.pm2/logs/openclaw-gateway-out.log last 15 lines:
0|openclaw | Gateway started on port 18789
0|openclaw | Telegram channel connected
0|openclaw | Health check: OK

Real-time monitoring:

pm2 monit

This opens an interactive dashboard showing CPU, memory, and logs in real-time.

PM2 Management Commands

Command	Purpose
`pm2 stop openclaw-gateway`	Stop the gateway
`pm2 restart openclaw-gateway`	Restart the gateway
`pm2 delete openclaw-gateway`	Remove from PM2
`pm2 flush`	Clear all log files
`pm2 logs --lines 100`	Show last 100 log lines

Option B: Oracle Cloud Free Tier

Oracle Cloud offers a genuinely free tier with substantial resources: up to 4 ARM CPUs and 24GB of RAM. This is more than enough for your AI Employee, and the cost is $0 forever (not a trial).

Why Oracle Cloud?

Provider	Free Tier	Specs	Notes
Oracle Cloud	Permanent	4 OCPU, 24GB RAM	ARM architecture
DigitalOcean	$200 trial	Various	Expires after 60 days
AWS	12 months	t2.micro	Very limited
Google Cloud	$300 trial	Various	Expires after 90 days
Hetzner	None	N/A	$4/mo cheapest

Oracle's Always Free tier is unique in offering substantial persistent compute at no cost.

Creating an Oracle Cloud Instance

Prerequisites:

Oracle Cloud account (signup at cloud.oracle.com/free)
SSH key pair
Approximately 30 minutes

Step 1: Create Instance

Log into Oracle Cloud Console
Navigate to Compute then Instances then Create Instance
Configure:
- Name: openclaw
- Image: Ubuntu 24.04 (aarch64)
- Shape: VM.Standard.A1.Flex (Ampere ARM)
- OCPUs: 2 (or up to 4)
- Memory: 12 GB (or up to 24 GB)
- Boot volume: 50 GB
- SSH key: Add your public key
Click Create

Note the public IP address when the instance is ready.

Common Issue: "Out of capacity" errors are common on the free tier. Try a different availability domain, or retry during off-peak hours (early morning works best).

Connecting and Setting Up

Step 2: Connect via SSH

ssh ubuntu@YOUR_PUBLIC_IP

Step 3: Update System

sudo apt update && sudo apt upgrade -y
sudo apt install -y build-essential

Build-essential is required for ARM compilation of some npm dependencies.

Step 4: Install Tailscale

Tailscale provides secure access without exposing ports to the internet:

curl -fsSL https://tailscale.com/install.sh | sh
sudo tailscale up --ssh --hostname=openclaw

Output:

To authenticate, visit:

  https://login.tailscale.com/a/1234567890abcdef

Success.

After authenticating in your browser, verify:

tailscale status

Output:

100.100.100.100  openclaw            linux   -
100.100.100.101  your-laptop         macOS   active; direct

From now on, you can connect via Tailscale instead of the public IP:

ssh ubuntu@openclaw

Step 5: Install OpenClaw

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

When prompted "How do you want to hatch your bot?", select "Do this later". We will configure manually.

Configuring for Remote Access

Step 6: Configure Gateway

# Bind only to localhost (Tailscale will provide access)
openclaw config set gateway.bind loopback

# Require token authentication
openclaw config set gateway.auth.mode token
openclaw doctor --generate-gateway-token

# Expose via Tailscale Serve (automatic HTTPS)
openclaw config set gateway.tailscale.mode serve
openclaw config set gateway.trustedProxies '["127.0.0.1"]'

Step 7: Enable as Persistent Service

# Enable user lingering (keeps services running after logout)
sudo loginctl enable-linger ubuntu

# Start the gateway service
systemctl --user restart openclaw-gateway

Verifying the Setup

# Check OpenClaw version
openclaw --version

Output:

OpenClaw v2025.2.3

# Check gateway status
systemctl --user status openclaw-gateway

Output:

● openclaw-gateway.service - OpenClaw Gateway
     Loaded: loaded (/home/ubuntu/.config/systemd/user/openclaw-gateway.service)
     Active: active (running) since Wed 2026-02-05 10:30:00 UTC
   Main PID: 12345 (openclaw)
      Tasks: 12 (limit: 24576)
     Memory: 156.0M
     CGroup: /user.slice/user-1000.slice/user@1000.service/openclaw-gateway.service
             └─12345 openclaw gateway run --port 18789

# Check Tailscale Serve status
tailscale serve status

Output:

https://openclaw.YOUR-TAILNET.ts.net (Funnel off)
|-- / proxy http://127.0.0.1:18789

# Test local response
curl http://localhost:18789

Output:

{"status":"ok","version":"2025.2.3"}

Accessing Your Remote Gateway

From any device on your Tailscale network, access the Control UI at:

https://openclaw.YOUR-TAILNET.ts.net/

Replace YOUR-TAILNET with your tailnet name (visible in tailscale status).

This works from:

Your laptop (at home or traveling)
Your phone (with Tailscale app installed)
Any other device on your tailnet

No SSH tunnel needed. Tailscale provides automatic HTTPS certificates and authentication.

Security Configuration

With Tailscale handling access, you can lock down the Oracle Cloud VCN (Virtual Cloud Network) to block all public traffic except Tailscale:

Go to Networking then Virtual Cloud Networks in Oracle Console
Click your VCN then Security Lists then Default Security List
Remove all ingress rules except: 0.0.0.0/0 UDP 41641 (Tailscale)

This blocks SSH, HTTP, HTTPS, and everything else at the network edge. Only Tailscale traffic reaches your instance.

Health Monitoring

Whether running locally with PM2 or remotely on Oracle Cloud, you want automatic recovery when things go wrong.

OpenClaw Health Commands

# Quick health check
openclaw health

Output:

Gateway: healthy
Channels:
  telegram: connected
  discord: disconnected (not configured)
Models:
  moonshot/kimi-k2.5: available
Memory: 156MB / 24GB
Uptime: 3d 14h 22m

# Comprehensive status
openclaw status --all

Output:

┌─────────────────────────────────────────────────────────────┐
│ OpenClaw Status                                              │
├─────────────────────────────────────────────────────────────┤
│ Gateway         Running on port 18789                        │
│ Uptime          3 days, 14 hours                            │
│ Sessions        47 active, 1,234 total                       │
│ Memory          156 MB                                       │
├─────────────────────────────────────────────────────────────┤
│ Channels                                                     │
│   Telegram      ✓ Connected (@YourBrandBot)                 │
│   WhatsApp      ✓ Connected (+1234567890)                   │
├─────────────────────────────────────────────────────────────┤
│ Models                                                       │
│   Primary       moonshot/kimi-k2.5 (available)              │
│   Fallback      gemini-2.0-flash (available)                │
└─────────────────────────────────────────────────────────────┘

# Follow logs in real-time
openclaw logs --follow

Monitoring Commands Reference

Command	Purpose
`openclaw health`	Quick status check
`openclaw status --all`	Comprehensive status
`openclaw logs --follow`	Real-time log output
`openclaw doctor`	Diagnose configuration issues
`openclaw security audit`	Check security posture

Cost Comparison Summary

Option	Monthly Cost	Pros	Cons
Local PM2	$0 (electricity)	Simple setup, fast	Computer must stay on
Oracle Free	$0	True 24/7, ARM capable	Setup complexity, capacity limits
DigitalOcean	$6	Easy UI, good docs	Monthly cost
Hetzner	$4	Cheapest paid, Docker-friendly	EU-focused
Linode	$5	Reliable, good support	Monthly cost

For most users, we recommend starting with PM2 locally to ensure everything works, then graduating to Oracle Cloud when you want true independence from your local machine.

Try With AI

Prompt 1: Deployment Analysis

I want my AI Employee running 24/7. Currently my computer is on about 12 hours
a day. Analyze my options: Should I use PM2 locally, deploy to Oracle Cloud,
or consider a paid VPS? What factors should I consider?

What you're learning: This prompt tests your AI Employee's understanding of deployment architecture. You are evaluating whether it can reason about your specific constraints (12-hour availability) and recommend appropriate solutions. Pay attention to whether it asks clarifying questions about your requirements (reliability needs, technical comfort, budget) or jumps to a recommendation.

Prompt 2: Health Check Discussion

Walk me through how to verify you're running correctly. What commands should
I use to check your health? What would indicate a problem? What would trigger
an automatic restart?

What you're learning: This prompt tests operational understanding. Your AI Employee should explain its own monitoring capabilities, demonstrating self-awareness of the infrastructure it runs on. A good response distinguishes between health checks (is it running?) and functionality checks (is it working correctly?).

Important: Deployment is advanced work. If you encounter issues, the troubleshooting commands in the Oracle Cloud documentation (referenced in this lesson) cover the most common problems. Start with PM2 locally to build confidence before attempting cloud deployment.

The Always-On Promise​

Option A: Local Always-On with PM2​

Installing PM2​

Creating an Ecosystem File​

Starting with PM2​

Persisting Across Reboots​

Monitoring PM2 Processes​

PM2 Management Commands​

Option B: Oracle Cloud Free Tier​

Why Oracle Cloud?​

Creating an Oracle Cloud Instance​

Connecting and Setting Up​

Configuring for Remote Access​

Verifying the Setup​

Accessing Your Remote Gateway​

Security Configuration​

Health Monitoring​

OpenClaw Health Commands​

Monitoring Commands Reference​

Cost Comparison Summary​

Try With AI​

Prompt 1: Deployment Analysis​

Prompt 2: Health Check Discussion​

The Always-On Promise

Option A: Local Always-On with PM2

Installing PM2

Creating an Ecosystem File

Starting with PM2

Persisting Across Reboots

Monitoring PM2 Processes

PM2 Management Commands

Option B: Oracle Cloud Free Tier

Why Oracle Cloud?

Creating an Oracle Cloud Instance

Connecting and Setting Up

Configuring for Remote Access

Verifying the Setup

Accessing Your Remote Gateway

Security Configuration

Health Monitoring

OpenClaw Health Commands

Monitoring Commands Reference

Cost Comparison Summary

Try With AI

Prompt 1: Deployment Analysis

Prompt 2: Health Check Discussion