Disclosure: We earn commissions from partner links. This doesn't affect our rankings. Learn more
Best VPS for OpenClaw in 2026
OpenClaw is an open-source, self-hosted gateway that connects chat apps to AI agents. The Gateway keeps sessions, routing, channels, memory, and tool access together on a machine or server you control.
A small Linux VPS is enough for the Gateway itself. The practical choice is between a managed OpenClaw deployment, a normal Linux install, and an optional Docker deployment. GPU hosting is relevant only when you also run a local model that needs a GPU.
Hostinger offers managed and VPS OpenClaw options
Starting at $3.99/mo. OpenClaw documents a managed one-click option and a Docker-based VPS path for Hostinger.
Get Hostinger VPS →OpenClaw on Hostinger - Video Tutorial
Watch the existing walkthrough, then compare its commands with the current official installation guide before deploying:
What is OpenClaw?
OpenClaw is the control plane between messaging channels and tool-capable AI agents. It can run on macOS, Linux, Windows, WSL2, or a VPS. The current stable release is v2026.9.1; v2026.8.1 is the release that the official notes also call OpenClaw 2.0.
A remote Gateway can stay available when your laptop is off, but it also holds powerful credentials and tools. Use one trust boundary per Gateway, keep access private, and run separate gateways for mutually untrusted users.
Official requirements and practical headroom
| Area | Official baseline | Practical headroom | What matters |
|---|---|---|---|
| RAM | 1 GB | 2 GB or more | Logs, media, multiple channels, browser automation, and node tools need more headroom |
| CPU | 1 vCPU | 1 to 2 vCPU | The Gateway itself does not need a GPU |
| Storage | About 500 MB | Extra room for state, logs, media, and workspaces | Actual usage depends on connected tools and retained files |
| Runtime | Node 22.22.3+, 24.15+, or 25.9+ | Node 26 | pnpm is needed only for a source checkout |
| Docker | Optional | Official prebuilt image on a small VPS | A local source-image build needs at least 6 GB RAM |
| GPU | Not required | Only for a local model workload that needs one | There is no universal OpenClaw VRAM minimum |
Best Hosting Providers for OpenClaw
Use a normal VPS for a hosted-model Gateway. Consider a GPU provider only when local inference is part of the same deployment. The price labels below are the existing site snapshot; verify current checkout terms with the provider.
| Provider | Type | RAM | Price | Use case | |
|---|---|---|---|---|---|
| Hostinger | VPS | 8 GB | $7.99/mo | Official managed or VPS path | Get Deal → |
| Hetzner | VPS | 8 GB | €5.59/mo | General Linux VPS | Get Deal → |
| vast.ai | GPU | 48 GB VRAM | $0.69/hr | Optional local model workload | Get Deal → |
| Vultr | VPS | 8 GB | $48/mo | General Linux VPS | Get Deal → |
| DigitalOcean | VPS | 8 GB | $48/mo | General Linux VPS | Get Deal → |
Safer setup path on a Hostinger VPS
1. Choose managed or self-managed
Hostinger offers an OpenClaw managed option and an OpenClaw on VPS path. Choose the managed route for provider-operated infrastructure and updates, or the VPS route when you want to administer the Linux host yourself.
2. Keep the Gateway private
Restrict the server firewall to SSH and other deliberate entry points. Do not expose the Gateway port directly to the public internet. Use an SSH tunnel or Tailscale for remote access.
3. Follow the maintained install flow
For Docker, use the official prebuilt image and maintained setup script. The script creates a Gateway token, runs onboarding, and starts the Gateway. Avoid unofficial container mirrors.
export OPENCLAW_IMAGE="ghcr.io/openclaw/openclaw:latest"
./scripts/docker/setup.sh4. Verify and audit
Confirm the installation with openclaw doctor, openclaw gateway status, and openclaw security audit. The Control UI normally uses port 18789 and should be reached through your private tunnel.
Check the current dedicated guide
Use our dedicated OpenClaw hosting site for more context, and verify commands against the official documentation linked below.
Visit open-claw-hosting.com →Official sources checked
- OpenClaw overview and architecture
- Current OpenClaw release notes
- Supported installation methods and Node versions
- Official VPS and VM requirements
- Official Hostinger deployment paths
- Gateway security and trust boundaries
- Docker on a VPS and private network exposure
Frequently Asked Questions
What are the minimum requirements for OpenClaw?
The official VPS baseline is 1 vCPU, 1 GB RAM, and about 500 MB of disk. OpenClaw recommends 1 to 2 vCPU and at least 2 GB RAM when logs, media, multiple channels, browser automation, or node tools need more headroom.
Can I run OpenClaw on Hostinger VPS?
Yes. The official OpenClaw documentation covers both a Hostinger managed one-click deployment and a Docker-based Hostinger VPS path. The existing KVM 4 price label on this page is $7.99/mo; verify current plan specifications and checkout terms before ordering.
Does OpenClaw require Docker or a GPU?
No. Docker is optional, and OpenClaw has no universal GPU requirement. GPU hosting matters only if you deliberately run a local model workload that needs one. A Gateway that calls a hosted model can run on a normal CPU VPS.
How much does it cost to host OpenClaw?
The existing site snapshot lists CPU-only hosting at $3.99/mo on Hostinger or €4.15/mo on Hetzner. It lists GPU hosting at $0.20-0.80/hour on vast.ai or RunPod depending on the GPU model. Verify current prices, taxes, renewal terms, and plan resources with the provider.
How should I access OpenClaw on a VPS?
Keep the Gateway port private. OpenClaw recommends SSH tunneling or Tailscale for remote access and warns against direct public exposure. Run openclaw security audit after configuration changes.