How to Host Your OpenClaw: VPS Comparison Guide
Where should you run OpenClaw? Compare DigitalOcean, Hetzner, AWS, Railway, and more. Find the best hosting for your AI agent.
You've built your OpenClaw agent. It works on your machine. It does something genuinely useful—maybe it monitors prices, handles customer support, scrapes data, manages workflows, or automates some tedious process that was eating hours of your week.
Now what?
"Works on my laptop" isn't a deployment strategy. Your agent needs to run 24/7, respond fast, and not crash when you close your MacBook lid. You need a server. And the moment you start looking at hosting options, you're hit with dozens of providers, confusing pricing tiers, and marketing pages that all say the same thing.
I've deployed OpenClaw agents across most of the major hosting providers at this point. Here's what actually matters, what each option is best for, and how to make the decision without overthinking it.
What OpenClaw Actually Needs From a Server
Before comparing anything, let's establish what we're working with.
OpenClaw agents are, at their core, persistent processes. They're not static websites. They're not simple API endpoints that spin up and die. They're long-running programs that need to maintain state, respond to triggers, and operate continuously. This means your hosting needs to support:
- Always-on compute — Your agent can't sleep. Cold starts kill the whole point.
- Sufficient RAM — Most OpenClaw agents run comfortably on 2-4GB RAM, but if you're running multiple agents or handling heavy context windows, 8GB is safer.
- Docker support — OpenClaw deploys cleanly via Docker. If a platform makes Docker hard, skip it.
- Persistent storage — Logs, state files, databases. You need disk that doesn't vanish on restart.
- Reasonable network — Your agent is making API calls, possibly serving webhooks. Bandwidth matters.
The baseline spec I'd recommend for a single OpenClaw agent: 2 vCPU, 4GB RAM, 80GB SSD. For running 2-5 agents on the same box, double the RAM.
With that established, let's look at the actual options.
The Contenders
I'm comparing five providers that cover the full spectrum: Hetzner, Vultr, DigitalOcean, AWS Lightsail, and Railway. Each occupies a different niche. There are others (Linode, Render, Fly.io), but these five cover 90% of use cases, and I'd rather go deep on five than shallow on ten.
The Pricing Reality
Here's what you'll actually pay monthly for that baseline spec (2 vCPU / 4GB RAM / 80GB SSD):
| Provider | Monthly Cost | Storage/Transfer | Hourly Billing? |
|---|---|---|---|
| Hetzner | ~$7-12 | 80GB NVMe, 20TB out | No (monthly) |
| Vultr | $24-28 | 80GB NVMe, 3TB out | Yes |
| DigitalOcean | $24 (Droplet) | 80GB SSD, 4TB out | Yes |
| AWS Lightsail | $20 | 80GB SSD, 3TB out | No (monthly) |
| Railway | $5-50+ (usage-based) | 8GB default | Yes |
That's not a typo on Hetzner. They're genuinely 2-3x cheaper than everyone else for raw compute. More on that in a second.
Hetzner: The Best Value, Period
Best for: Budget-conscious deployments, high-performance agents, anyone comfortable with basic server management
Hetzner's CX32 plan gives you 4 vCPU, 8GB RAM, and 80GB NVMe storage for roughly €11/month (~$12). Their entry-level CX22 is €6.49/month for 2 vCPU and 4GB RAM. Both come with 20TB of outbound transfer, which is absurd generosity compared to everyone else.
The hardware is modern AMD EPYC. The NVMe storage is fast. The network is solid. For the price, nothing else comes close.
Here's how to deploy an OpenClaw agent on Hetzner:
# SSH into your Hetzner VPS after creation
ssh root@your-server-ip
# Install Docker
curl -fsSL https://get.docker.com | sh
# Create a directory for your agent
mkdir -p /opt/openclaw && cd /opt/openclaw
# Create your docker-compose.yml
cat > docker-compose.yml << 'EOF'
version: '3.8'
services:
openclaw-agent:
image: openclaw/agent:latest
container_name: openclaw-agent
restart: always
env_file: .env
volumes:
- ./data:/app/data
- ./logs:/app/logs
ports:
- "8080:8080"
EOF
# Create your environment file
cat > .env << 'EOF'
OPENCLAW_API_KEY=your-api-key-here
OPENCLAW_AGENT_CONFIG=./data/agent-config.json
LOG_LEVEL=info
EOF
# Launch it
docker compose up -d
# Verify it's running
docker compose logs -f
Set up a basic firewall while you're at it:
# Allow SSH and your agent's port
ufw allow 22/tcp
ufw allow 8080/tcp
ufw enable
That's it. Your OpenClaw agent is running on hardware that costs less than two coffees a month.
The downsides: Hetzner's datacenters are EU-focused (Falkenstein, Nuremberg, Helsinki) with US options in Ashburn and Hillsboro. If your users or API endpoints are in Asia-Pacific, latency will be noticeable. The UI is functional but no-frills. There's no managed database offering worth mentioning. And their abuse policy is strict—they don't mess around with DMCA complaints.
When to pick Hetzner: You want maximum performance per dollar, you're comfortable with SSH and basic Linux administration, and your latency requirements are satisfied by EU or US-East locations.
DigitalOcean: The Comfortable Middle Ground
Best for: Teams, developers who want managed services, people who value good documentation
DigitalOcean is the Honda Accord of cloud hosting. Nothing about it is the absolute best, but nothing about it is bad either. It just works, and the experience of using it is genuinely pleasant.
A 2 vCPU / 4GB RAM Droplet runs $24/month. You get 80GB SSD and 4TB of transfer. The control panel is the best in the industry for readability. The documentation reads like it was written by humans who actually use the product (because it was).
What makes DigitalOcean interesting for OpenClaw is the App Platform. If you don't want to manage a server at all, you can connect your OpenClaw agent's Git repository and deploy automatically on every push:
# .do/app.yaml — DigitalOcean App Platform spec
name: openclaw-agent
services:
- name: agent
github:
repo: your-username/your-openclaw-agent
branch: main
deploy_on_push: true
dockerfile_path: Dockerfile
instance_count: 1
instance_size_slug: professional-s
envs:
- key: OPENCLAW_API_KEY
value: your-api-key
type: SECRET
- key: LOG_LEVEL
value: info
health_check:
http_path: /health
Push to main, and it deploys. No SSH, no Docker commands, no server maintenance. App Platform starts around $12/month for basic instances, scaling up as your needs grow.
DigitalOcean also offers managed PostgreSQL ($15/month starting), Spaces object storage ($5/month for 250GB), and a built-in monitoring stack. If your OpenClaw agent needs a database for state management—and many do—having a managed Postgres instance next to your agent with single-digit millisecond latency is genuinely nice.
The downsides: You're paying a 2x premium over Hetzner for essentially the same compute. Egress beyond 4TB gets expensive. And while Droplets are fine for most workloads, the CPU performance per dollar doesn't match Hetzner's EPYC hardware.
When to pick DigitalOcean: You want a balance of ease-of-use and control, you value great docs, or you need managed databases and object storage alongside your OpenClaw agent without stitching together multiple providers.
AWS Lightsail: The Gateway Drug
Best for: Teams already in the AWS ecosystem, agents that might need to scale into full AWS services
Lightsail is AWS for people who don't want to deal with AWS. The $20/month plan gets you 2 vCPU, 4GB RAM, 80GB SSD, and 3TB transfer with flat, predictable pricing. No surprise bills. No accidentally leaving a NAT gateway running at $32/month (ask me how I know).
The real advantage of Lightsail is the upgrade path. Your OpenClaw agent starts simple. Then you need a queue system—add SQS. Then you need event triggers—add Lambda. Then you need a proper database—add RDS. Lightsail lets you start simple and gradually adopt the full AWS toolkit without migrating providers.
# Install the AWS CLI and create a Lightsail instance
aws lightsail create-instances \
--instance-names openclaw-agent-01 \
--availability-zone us-east-1a \
--blueprint-id ubuntu_22_04 \
--bundle-id medium_3_0 \
--user-data '#!/bin/bash
curl -fsSL https://get.docker.com | sh
usermod -aG docker ubuntu
mkdir -p /opt/openclaw
echo "Setup complete" > /opt/openclaw/init.log'
# Attach a static IP
aws lightsail allocate-static-ip --static-ip-name openclaw-ip
aws lightsail attach-static-ip \
--static-ip-name openclaw-ip \
--instance-name openclaw-agent-01
The downsides: You're in the AWS ecosystem now. That's a feature and a bug. Migration out of AWS is always harder than migration in. Lightsail's customization options are more limited than full EC2. And while $20/month is predictable, it's not cheap for what you get compared to Hetzner.
When to pick Lightsail: You're already using AWS for other things, or you anticipate needing AWS services (S3, Lambda, SQS, CloudFront) as your OpenClaw agent grows in complexity.
Vultr: The Global Option
Best for: Agents that need to be close to users worldwide, high-frequency compute needs
Vultr has 30+ datacenter locations spanning every continent except Antarctica. If your OpenClaw agent is serving users in São Paulo, Johannesburg, Mumbai, and Tokyo, Vultr is the only provider on this list that can put a server within reasonable latency of all of them.
Their High-Frequency Compute plans use high-clock-speed CPUs (3GHz+ NVMe-backed), which matters for CPU-intensive OpenClaw agents doing heavy processing or running local models alongside the agent logic.
Pricing is competitive with DigitalOcean—$24-28/month for the baseline spec—with hourly billing so you can spin up testing instances without commitment.
The downsides: Bandwidth caps are tighter than Hetzner or DigitalOcean (3TB), and overage fees are $0.01/GB. The UI is good but not as polished as DigitalOcean's. Block storage is cheap ($10/TB) but not available in all regions.
When to pick Vultr: Geographic diversity is a hard requirement, or you need high-clock-speed CPUs for compute-heavy OpenClaw agents.
Railway: The "I Don't Want to Manage Servers" Option
Best for: Prototyping, developers who want Heroku-style simplicity, multi-service architectures
Railway is the most opinionated option here, and that's its strength. Connect a GitHub repo. Railway detects your Dockerfile (or builds from a Nixpack). It deploys. Need Postgres? Click a button. Redis? Another button. Environment variables? Set them in the dashboard.
# Install Railway CLI
npm install -g @railway/cli
# Login and initialize
railway login
railway init
# Link your OpenClaw project
railway link
# Add a Postgres database with one command
railway add --plugin postgresql
# Deploy
railway up
The Hobby plan gives you $5/month in credits, which covers very light usage. A continuously-running OpenClaw agent will cost $20-50/month depending on resource consumption, since Railway bills per vCPU-second and GB-RAM-second.
The downsides, and they're significant for OpenClaw: Railway is designed for web services, not persistent agents. It can sleep idle containers. Cold starts can introduce delays. You don't get SSH access by default. Persistent storage requires explicitly configured volumes (only 8GB default). And usage-based pricing means your bill is unpredictable—a bug that causes a CPU spike could cost real money.
When to pick Railway: You're prototyping your OpenClaw agent and want it online in five minutes. You're running a lightweight agent that responds to webhooks rather than running continuous loops. Or you simply refuse to touch a terminal for server administration.
The Decision Framework
Stop overthinking this. Here's the flowchart:
"I want the cheapest option and I know Linux" → Hetzner. Start with CX22 at €6.49/month. Deploy Docker. Done.
"I want easy but not braindead, with room to grow" → DigitalOcean. Droplets for control, App Platform for convenience. $24/month.
"I'm already in AWS / I'll need Lambda, SQS, etc." → AWS Lightsail. $20/month, seamless upgrade path to full AWS.
"My users are everywhere on the planet" → Vultr. 30+ locations. $24/month.
"I just want to prototype this thing right now" → Railway. Free to start. Deployed in minutes.
"I'm running 5+ OpenClaw agents and need serious compute" → Hetzner Dedicated or Vultr Bare Metal. $50-100/month for hardware that would cost $300+ elsewhere.
Post-Deployment: Don't Forget This Stuff
Whichever provider you choose, do these things immediately after your OpenClaw agent is running:
Set up monitoring. UptimeRobot (free) will ping your agent every 5 minutes and alert you when it goes down. Betterstack is worth the $10/month if you want proper incident management.
Configure automatic restarts. The restart: always flag in Docker Compose handles this. If your agent crashes, Docker will bring it back.
Set up log rotation. OpenClaw agents can generate substantial logs. Without rotation, you'll fill your disk:
# /etc/docker/daemon.json
{
"log-driver": "json-file",
"log-opts": {
"max-size": "10m",
"max-file": "3"
}
}
Enable automatic security updates:
sudo apt install unattended-upgrades
sudo dpkg-reconfigure -plow unattended-upgrades
Back up your agent's state. If your OpenClaw agent stores anything in its data directory, snapshot it regularly. Hetzner snapshots are €0.012/GB/month. DigitalOcean charges $0.05/GB/month. Both are trivial compared to losing your agent's learned state or configuration.
The Bottom Line
For most people deploying their first OpenClaw agent, I'd say start with Hetzner if you're comfortable with servers, or DigitalOcean if you want guardrails. Both will serve you well for months or years before you need to think about switching.
The worst thing you can do is spend three days "researching hosting" instead of deploying. Pick one, get your OpenClaw agent running, and move on to the work that actually matters—making your agent smarter, more reliable, and more useful.
Your agent isn't doing anything sitting on your laptop. Get it on a server today.