How to Host OpenClaw in the Cloud (So It Keeps Running After You Close Your Laptop)

Host OpenClaw on a cloud VPS using Docker with 4GB+ RAM and 2+ CPU cores for 24/7 operation. Deploy using docker run -d -p 3000:3000 openclaw/openclaw on DigitalOcean ($6/mo), AWS Lightsail ($10/mo), or Hetzner ($5/mo). Configure environment variables for API keys, set up systemd for auto-restart, and secure with SSH key authentication and firewall rules.

What Is OpenClaw and Why Do People Run It?

OpenClaw is an open-source autonomous AI agent platform with over 140,000 GitHub stars. It executes complex tasks by breaking them into steps, writing code, running commands, and iterating until completion.

Unlike traditional AI assistants that answer questions, OpenClaw takes action. It builds apps, fixes bugs, analyzes data, and automates workflows without constant supervision.

The platform runs locally by default. When your laptop sleeps or closes, OpenClaw stops. For long-running tasks or team collaboration, you need cloud hosting.

The Problem: Your Computer Has to Stay Awake

Local OpenClaw deployment has limitations:

• Stops when your computer sleeps. Tasks fail mid-execution.
• No remote access. You can't check progress from another device.
• Single-user only. Team members can't share agents or workflows.
• Resource intensive. Running AI models drains laptop battery and CPU.

Cloud hosting solves these problems. Your agent runs 24/7 on dedicated infrastructure, accessible from anywhere, with shared team access.

What Are the Cloud Hosting Options?

You have two paths: self-host on a VPS or use a managed platform.

Self-hosting gives you full control. You configure the server, manage updates, handle security, and pay only for compute resources.

Managed platforms handle infrastructure, security, and updates automatically. You focus on using the agent, not maintaining servers.

How to Self-Host OpenClaw on a VPS

Self-hosting requires a Linux server, Docker, and basic command-line skills. Here's the complete setup.

Step 1: Choose a VPS Provider

Select a provider based on pricing and location:

OpenClaw requires minimum 4GB RAM and 2 CPU cores for stable operation. The $5 Hetzner option is the most cost-effective for full performance.

For lighter usage, a 2GB instance ($10-12/mo) works but may timeout on large tasks.

Step 2: Provision and Access Your Server

Create a new instance with Ubuntu 22.04 LTS. Add your SSH public key during setup.

Connect to your server:

ssh root@YOUR_SERVER_IP

Update system packages:

apt update && apt upgrade -y

Step 3: Install Docker

Install Docker Engine:

curl -fsSL https://get.docker.com -o get-docker.sh
sh get-docker.sh

Verify installation:

docker --version

Enable Docker to start on boot:

systemctl enable docker

Step 4: Configure Environment Variables

Create a directory for OpenClaw:

mkdir -p /opt/openclaw
cd /opt/openclaw

Create an environment file:

nano .env

Add your configuration:

ANTHROPIC_API_KEY=your_anthropic_key
OPENAI_API_KEY=your_openai_key
WORKSPACE_DIR=/workspace
PORT=3000

Save and exit (Ctrl+X, Y, Enter).

Step 5: Run OpenClaw with Docker

Pull the latest OpenClaw image:

docker pull openclaw/openclaw:latest

Run the container:

docker run -d \
  --name openclaw \
  --restart unless-stopped \
  -p 3000:3000 \
  -v /opt/openclaw/workspace:/workspace \
  --env-file /opt/openclaw/.env \
  openclaw/openclaw:latest

Check container status:

docker ps

View logs:

docker logs -f openclaw

Step 6: Configure Firewall and Security

Set up UFW firewall:

ufw allow OpenSSH
ufw allow 3000/tcp
ufw enable

Disable password authentication:

nano /etc/ssh/sshd_config

Set PasswordAuthentication no. Restart SSH:

systemctl restart sshd

Step 7: Set Up SSL with Nginx (Optional)

For HTTPS access, install Nginx and Certbot:

apt install nginx certbot python3-certbot-nginx -y

Configure Nginx reverse proxy:

nano /etc/nginx/sites-available/openclaw

Add configuration:

server {
    listen 80;
    server_name your-domain.com;

    location / {
        proxy_pass http://localhost:3000;
        proxy_http_version 1.1;
        proxy_set_header Upgrade $http_upgrade;
        proxy_set_header Connection 'upgrade';
        proxy_set_header Host $host;
        proxy_cache_bypass $http_upgrade;
    }
}

Enable site and get SSL certificate:

ln -s /etc/nginx/sites-available/openclaw /etc/nginx/sites-enabled/
certbot --nginx -d your-domain.com

Step 8: Monitor and Update

Check OpenClaw health:

docker stats openclaw

Update to latest version:

docker pull openclaw/openclaw:latest
docker stop openclaw
docker rm openclaw
docker run -d \
  --name openclaw \
  --restart unless-stopped \
  -p 3000:3000 \
  -v /opt/openclaw/workspace:/workspace \
  --env-file /opt/openclaw/.env \
  openclaw/openclaw:latest

Set up automatic backups of /opt/openclaw/workspace:

crontab -e

Add daily backup:

0 2 * * * tar -czf /backup/openclaw-$(date +\%Y\%m\%d).tar.gz /opt/openclaw/workspace

What Are the Security Risks of Self-Hosting?

Exposing an AI agent to the internet creates attack surface.

API Key Exposure

OpenClaw stores API keys in environment variables. If an attacker gains container access, they can extract keys and rack up charges on your accounts.

Mitigation:

• Use read-only API keys where possible
• Set spending limits on OpenAI/Anthropic accounts
• Rotate keys monthly
• Never commit .env files to version control

Code Execution Vulnerabilities

OpenClaw executes arbitrary code. Malicious prompts could exploit this to run unintended commands.

CVE-2024-47068 affected similar AI agent platforms in October 2024, allowing remote code execution through prompt injection.

Mitigation:

• Run OpenClaw in isolated containers with limited filesystem access
• Use non-root users inside containers
• Implement network segmentation
• Monitor logs for suspicious activity

Skill and Plugin Risks

OpenClaw supports custom skills. Unverified skills can contain malicious code.

Mitigation:

• Review skill source code before installation
• Use only official or vetted community skills
• Run skills in sandboxed environments
• Maintain an allowlist of approved skills

How Does Managed Hosting Compare?

Managed platforms eliminate infrastructure maintenance but reduce flexibility.

What You Get with Managed Platforms

Managed AI agent platforms handle:

• Server provisioning and scaling
• Security patches and updates
• SSL certificates and domain configuration
• Team collaboration features
• Built-in authentication and access control
• Monitoring and alerting

You access the agent through a web interface without touching Docker or Linux commands.

Cost Comparison

Self-hosting is cheaper for single users comfortable with server administration. Managed platforms save time and reduce security risk for teams.

When to Choose Self-Hosting

Choose self-hosting if you:

• Want maximum control over infrastructure
• Have DevOps expertise or time to learn
• Need custom configurations not supported by platforms
• Run sensitive workloads that can't use shared infrastructure

When to Choose Managed Hosting

Choose managed platforms if you:

• Want to start immediately without setup
• Lack server administration experience
• Need team collaboration features
• Prefer predictable costs without surprise server bills

What About Managed Alternatives to OpenClaw?

If you like OpenClaw's autonomous capabilities but want cloud infrastructure handled for you, several platforms offer similar functionality with better team features.

Duet provides an always-on AI agent that runs in the cloud with no server setup required. Unlike self-hosted OpenClaw, you get team workspaces, built-in security, and verified skills that work out of the box. Your agents run continuously on secure infrastructure with automatic scaling, so you never worry about server configuration, Docker updates, or API key management.

The platform includes collaboration features OpenClaw lacks: shared conversation history, team skill libraries, and real-time multi-user access. You can @mention the AI agent from any device, and it handles long-running tasks in the background while you close your laptop.

For teams moving from local development to production AI workflows, Duet bridges the gap between DIY open source tools and fully managed infrastructure. Learn more at duet.so.

How Do You Migrate from Local to Cloud?

Moving an existing OpenClaw setup to the cloud preserves your configuration and conversation history.

Step 1: Export Local Data

On your local machine, locate the OpenClaw workspace directory. Default locations:

• Linux/macOS: ~/.openclaw/workspace
• Windows: C:\Users\YourName\.openclaw\workspace

Archive the workspace:

tar -czf openclaw-backup.tar.gz ~/.openclaw/workspace

Step 2: Transfer to VPS

Copy the archive to your server:

scp openclaw-backup.tar.gz root@YOUR_SERVER_IP:/opt/openclaw/

SSH into the server and extract:

cd /opt/openclaw
tar -xzf openclaw-backup.tar.gz

Step 3: Update Configuration

Copy environment variables from local .env to server /opt/openclaw/.env.

Verify API keys work by testing with curl:

curl https://api.anthropic.com/v1/models \
  -H "x-api-key: $ANTHROPIC_API_KEY"

Step 4: Start Cloud Instance

Run Docker container with migrated workspace:

docker run -d \
  --name openclaw \
  --restart unless-stopped \
  -p 3000:3000 \
  -v /opt/openclaw/workspace:/workspace \
  --env-file /opt/openclaw/.env \
  openclaw/openclaw:latest

Access via http://YOUR_SERVER_IP:3000 and verify conversation history loaded.

What Are the Ongoing Maintenance Requirements?

Cloud-hosted OpenClaw requires regular maintenance.

Weekly Tasks

• Review Docker logs for errors: docker logs openclaw --tail 100
• Check disk usage: df -h
• Monitor API spending in Anthropic/OpenAI dashboards

Monthly Tasks

• Update OpenClaw to latest version (see Step 8 above)
• Rotate API keys
• Review and clean old workspace files
• Check security advisories for dependencies

Quarterly Tasks

• Audit installed skills for updates
• Review firewall rules
• Test backup restoration process
• Evaluate server sizing (upgrade if hitting resource limits)

Automation

Reduce manual work with systemd timers or cron jobs:

# /etc/cron.weekly/openclaw-update
#!/bin/bash
docker pull openclaw/openclaw:latest
docker stop openclaw
docker rm openclaw
docker run -d --name openclaw --restart unless-stopped -p 3000:3000 \
  -v /opt/openclaw/workspace:/workspace \
  --env-file /opt/openclaw/.env \
  openclaw/openclaw:latest

How Do You Handle High Availability?

Single-server deployments fail when the server goes down. High availability requires redundancy.

Load Balancer Setup

Deploy multiple OpenClaw instances behind a load balancer:

1. Provision 2-3 identical VPS instances
2. Set up shared storage (NFS or S3) for workspace directory
3. Configure Nginx or Traefik as load balancer
4. Point DNS to load balancer IP

Cost increases 2-3x but eliminates single point of failure.

Database Replication

If OpenClaw stores state in a database, configure primary-replica replication:

# On primary server
docker run -d \
  --name postgres-primary \
  -e POSTGRES_PASSWORD=secure_password \
  -e POSTGRES_USER=openclaw \
  -p 5432:5432 \
  postgres:15

Configure replicas to stream from primary. Update OpenClaw config to use read replicas for queries.

Health Monitoring

Set up uptime monitoring with UptimeRobot or Healthchecks.io:

1. Create HTTP health check endpoint
2. Configure alert if endpoint returns non-200 status
3. Set notification to email/Slack on failure

Most teams don't need HA for development use cases. Production workflows justify the added complexity.

Frequently Asked Questions

How much does it cost to host OpenClaw in the cloud?

Hosting OpenClaw on a VPS costs $5-20/month for the server plus API usage fees. A Hetzner CAX11 ($5/mo) with 4GB RAM handles most workloads. Add $10-50/mo for Anthropic/OpenAI API calls depending on usage. SSL certificates via Let's Encrypt are free.

Can I run OpenClaw on AWS free tier?

AWS free tier includes 750 hours of t2.micro instances with 1GB RAM. This is insufficient for OpenClaw, which requires minimum 4GB RAM. You'll need a t3.medium ($30/mo) or t3.large ($60/mo) instance for reliable performance.

What's the easiest way to deploy OpenClaw?

The easiest deployment method is using Docker on a managed VPS. Providers like DigitalOcean and Hetzner offer one-click Docker images. After provisioning, run a single docker run command with your API keys. Setup takes 15-30 minutes including SSL configuration.

How do I secure API keys on a cloud server?

Store API keys in environment files outside the container with restricted permissions: chmod 600 /opt/openclaw/.env. Use read-only API keys where possible. Rotate keys monthly. Enable audit logging to detect unauthorized access. Consider using secret management tools like Vault for production deployments.

Does OpenClaw work with custom AI models?

Yes, OpenClaw supports custom models through OpenAI-compatible API endpoints. Configure the OPENAI_API_BASE environment variable to point to your model provider (e.g., Ollama, LocalAI, or self-hosted LLaMA). Performance depends on model capability and hosting infrastructure.

Can multiple people use the same OpenClaw instance?

OpenClaw's default setup is single-user. Multiple people accessing the same instance share conversation history and workspace files, which creates confusion. For team use, deploy separate instances per user or use a managed platform with built-in multi-user support.

What happens if my server crashes?

If you configured --restart unless-stopped, Docker automatically restarts the OpenClaw container when the server reboots. In-progress tasks may fail and require manual restart. Implement health monitoring and alerting to detect outages quickly. Regular workspace backups prevent data loss.

Related Reading

For more on AI agent deployment and automation:

• How to Run Claude Code in the Cloud - Deploy Claude's AI coding assistant on remote servers
• OpenClaw vs. Managed AI Agent Platforms - Compare self-hosted and managed solutions
• How to Set Up a 24/7 AI Agent - Configure always-on autonomous agents
• How to Build and Ship an Internal Tool in a Day Using AI - Rapid prototyping with AI agents
• How to Scrape, Analyze, and Monitor Any Website - Automate web data collection
• Claude Code vs. Cursor vs. Codex - Compare AI coding assistants
• How to Build and Deploy a Web App Using Only AI - End-to-end AI development workflows

Conclusion

Hosting OpenClaw in the cloud transforms a local development tool into a 24/7 autonomous agent. Self-hosting on a VPS costs $5-20/month and gives you full control, but requires ongoing security maintenance and server administration.

For teams or those who want to skip infrastructure management, managed platforms handle deployment, security, and scaling automatically. The choice depends on your technical comfort level, budget, and collaboration requirements.

Start with a basic VPS deployment to understand OpenClaw's cloud hosting needs. As usage grows, migrate to high-availability setups or managed platforms that offer better team features and reduced operational overhead.