May 12, 2026 · 7 min read

Where Should You Run OpenClaw? (Please Don't Say Your Laptop)

VPS vs. repurposed hardware vs. Proxmox LXC vs. the daily driver — a practical breakdown of the four options.

When you first hear about OpenClaw — an open-source AI agent platform that runs your own agents on your own hardware — the natural instinct is to git clone it onto whatever machine is in front of you and hit openclaw start.

I get it. I did it too.

But running a 24/7 AI agent gateway on your personal computer is like running a web server on your phone. It works until it doesn’t, and when it breaks, it breaks your whole day.

Here’s a practical breakdown of the four real options, based on things I’ve actually tried.

Option 1: Your Daily Driver PC (The “Don’t”)

This is the path of least resistance. You install OpenClaw on your main laptop or desktop, start it up, and everything works. For a while.

The problems show up fast:

  • Your computer needs to stay on. Every reboot, every sleep cycle, every time you close the lid to go to a coffee shop — your agents go offline. Telegram messages queue up. Scheduled tasks miss their window. If you’re running heartbeats or cron jobs, they just don’t fire.

  • Resource contention. LLM inference is hungry. Even just proxying API calls and running agent logic, OpenClaw chews through CPU and RAM. If you’re compiling code, rendering video, or playing a game, something’s getting starved.

  • Security exposure. Your agent gateway is a network-accessible service. If you expose it to the internet (to connect Telegram, Signal, or a web app), you’re now running a production-adjacent service on your personal machine. One misconfigured endpoint and your local files are reachable.

  • Uptime is a joke. Your personal computer is not a server. It reboots for updates. It goes to sleep. It travels with you. Your agents should be running while you’re asleep, on a plane, or in a meeting.

Verdict: Fine for testing. Bad for anything real. Don’t do it.

Option 2: A Repurposed Old Laptop or PC

This one has sentimental appeal. That old ThinkPad in the closet? The desktop you replaced last year? It can be a server now.

I ran OpenClaw on an old Intel NUC for a month. Here’s the reality:

Pros:

  • It’s free if you already have it
  • Dedicated hardware — nothing else fighting for resources
  • You can leave it on 24/7
  • Full control over the machine

Cons:

  • Power draw. An old desktop idling can pull 50-100W. Running 24/7/365, that’s $50-150/year in electricity alone. A VPS is cheaper.
  • Heat and noise. Old hardware wasn’t built for silent operation. If it’s in your living space, you’ll hear it.
  • Reliability. Old hardware fails. Hard drives die. Power supplies die. Fans seize. You’re one component failure away from your agents going dark until you notice and fix it.
  • No redundancy. If the machine goes down, you’re rebuilding from scratch or restoring a backup. There’s no vendor handling the hardware layer.
  • Remote management is painful. No IPMI, no out-of-band console. If the OS hangs, you need physical access or a creative workaround.

Best for: Hobbyists who enjoy tinkering and already have suitable hardware collecting dust. Worst case, you break even on cost vs. a VPS after a year or two.

Option 3: Proxmox LXC (How I Run It)

This is my current setup. I have a Proxmox hypervisor running a few LXCs, and OpenClaw lives in its own lightweight container.

Why this works well:

  • Minimal overhead. An LXC is not a VM — it shares the host kernel. The resource cost is negligible. OpenClaw runs with the system resources it needs and nothing more.
  • Snapshots and backups. Proxmox can snapshot an LXC in seconds. Before any upgrade or config change, I snapshot. If something breaks, I roll back in ten seconds. This alone is worth the setup effort.
  • Resource isolation. I give the OpenClaw container 2 cores and 4GB RAM. It can’t OOM the rest of my services. It can’t eat CPU that my Pi-hole or NAS needs.
  • Network isolation. The container lives on its own VLAN. The gateway can reach the internet (for LLM APIs), but I can restrict what it can reach on my local network.
  • 24/7 uptime. The Proxmox host stays on. The LXC auto-starts. I reboot the host once a quarter for kernel updates and plan it. My agents don’t go down because I closed my laptop.

Cons:

  • You need Proxmox first. This assumes you already run or are willing to set up a hypervisor. It’s free, but it’s a learning curve.
  • Single point of failure. The Proxmox host is still one machine. If it dies, all your LXCs die with it. This is fine for a homelab; it’s not fine for production.
  • No external access without setup. If your agents need to be reachable from the internet (Telegram webhooks, web apps), you need a reverse proxy, DNS, and probably a Cloudflare tunnel or VPN. This isn’t hard, but it’s additional config.

Verdict: The sweet spot for homelab setups. If you already run Proxmox, this is the obvious choice. If you don’t, a VPS is simpler.

Option 4: A VPS (The “Just Works” Option)

A $5-15/month VPS from any provider — DigitalOcean, Hetzner, Linode, Vultr — running Ubuntu, OpenClaw installed, agents configured.

Pros:

  • True 24/7 uptime. The provider handles hardware, power, networking. Your agents run while you sleep.
  • Static IP / DNS. Easy to configure webhooks. Easy to point a domain at it. No dynamic DNS, no NAT traversal.
  • Snapshots for backup. Most providers offer VM snapshots. Easy rollback.
  • No hardware to babysit. No fans to clean, no drives to replace, no power bills.
  • Scales trivially. Need more RAM? Click a button. Need a bigger disk? Click a button.

Cons:

  • Monthly cost. $5-15/month adds up. $60-180/year. It’s cheap, but it’s not free.
  • You don’t own the hardware. Data leaves your home network. For sensitive agent memory (personal info, business data), you’re trusting the provider.
  • Latency. If you’re running agents that need to act locally (smart home, local file system), a remote VPS adds latency and complexity.
  • Less fun. A VPS is boring. You SSH in, install, configure, and it just works. Tinkerers may prefer the homelab route.

Verdict: The safest recommendation for most people. It’s the default for good reason.

The Comparison

Factor Daily Driver Old Hardware Proxmox LXC VPS
Cost Free Free (+ power) Free (if you have Proxmox) $5-15/mo
Uptime Terrible Good Great Best
Security Risky Good Great Good
Setup effort None Medium High Low
Maintenance None High Medium None
Remote access Bad Bad Medium Built-in
Fun factor Low High High Low

My Recommendation

If you’re just starting: Get a $6 Hetzner VPS. Install Ubuntu. Install OpenClaw. Go. It’ll take 20 minutes and you’ll never think about it again.

If you already have a homelab: Use a Proxmox LXC. Snapshot before every change. Sleep well.

If you have old hardware and enjoy the hobby: Go for it. Just track your power draw and have a backup plan for when the hardware dies.

Don’t use your daily driver. Seriously. Your agents should be running right now, not when you’re at your desk.


OpenClaw is open source on GitHub. If you want pre-configured agents that work out of the box — an AI Dungeon Master, a fitness coach, a business assistant — they’re available on Gumroad as self-contained packages.


← Back to all posts