Custom Server BYOS

Deploy OpenClaw on Any Machine

Cloud VPS, colocation rack, office server, or a box under your desk. If it has SSH and runs Linux, OpenClaw runs on it.

Connect Any Server in Three Steps

Provide SSH credentials and OpenClawHosting handles the rest

01

Enter SSH Details

Provide the server’s IP, SSH port, and authentication method (key or password). We verify connectivity and OS compatibility automatically.

02

Validation & Install

OpenClawHosting checks CPU, RAM, and disk against minimum requirements, then installs Docker, OpenClaw, and all dependencies.

03

Configure & Launch

Add your AI provider API key, connect messaging platforms, and your agent goes live. The entire process takes about five minutes.

Maximum Flexibility

No vendor restrictions, no proprietary APIs, no limits on what you can install

Any Provider, Any Hardware

OVH, Scaleway, Oracle Cloud, Google Cloud, IBM, Alibaba—or a rack-mounted box in your closet. If it runs Linux and accepts SSH, OpenClaw runs on it.

On-Premise & Colocation

Deploy in your office, data center cage, or home lab. Keep hardware physically close and eliminate reliance on third-party cloud providers entirely.

Air-Gapped Mode

Pair OpenClaw with a local language model (Ollama, llama.cpp) and an internal messaging bridge. No internet connection required after initial setup.

Unrestricted Configuration

Custom kernels, non-standard port ranges, internal DNS, VPN tunnels—nothing is off limits when you control the entire stack.

Works With Any Provider

OVHcloudScalewayOracle CloudGoogle CloudIBM CloudAlibaba CloudContaboHostinger VPSKamateraUpCloudCherry ServersLatitude.shAny colo / bare metal

Hardware Requirements

Minimum specs to run OpenClaw and recommended specs for production

SpecMinimumRecommended
Operating SystemUbuntu 22.04+ or Debian 11+Ubuntu 24.04 LTS
CPU2 cores (x86_64 or ARM64)4+ cores
RAM2 GB4 GB (16 GB for local models)
Disk20 GB SSD50 GB NVMe
NetworkPublic IP or reverse proxy + SSHStatic IP with ports 80, 443 open
GPU (optional)N/ANVIDIA with 8 GB+ VRAM for Ollama

Local Model Inference

Running models like Llama 3 via Ollama requires an NVIDIA GPU with at least 8 GB VRAM, 16 GB system RAM, and 50 GB free disk space. CPU-only inference is possible with smaller quantized models but significantly slower.

Custom Server FAQ

What do I need to provide for custom server setup?

An IP address (or hostname), SSH port, a user with sudo privileges, and either an SSH key or password. OpenClawHosting connects over SSH, validates the server meets requirements, installs dependencies, and deploys OpenClaw.

Can I run OpenClaw on ARM64 hardware like Raspberry Pi?

OpenClaw supports ARM64 Linux servers. A Raspberry Pi 4 with 4 GB RAM or better can run a single agent for light workloads. ARM-based cloud instances (AWS Graviton, Oracle Ampere) are also supported and often cheaper than x86.

How does air-gapped deployment work?

During initial setup, OpenClawHosting downloads all dependencies and the OpenClaw package. After that, disconnect the server from the internet. Use a local language model for inference and an internal messaging bridge (e.g., Mattermost or Matrix) for communication. Updates require a brief reconnection or offline package transfer.

Can I run OpenClaw alongside other services?

Yes. OpenClaw runs in Docker containers with isolated networking. It coexists with web servers, databases, and other applications on the same machine. We recommend dedicating at least 2 GB RAM to OpenClaw to avoid contention.

What if my server uses a non-standard Linux distribution?

OpenClawHosting officially supports Ubuntu 22.04+ and Debian 11+. Other distributions (Fedora, Rocky Linux, Arch) may work but are not tested. If your distribution supports Docker and systemd, OpenClaw will likely run without issues.

What are the costs with a custom server?

OpenClawHosting starts at $29/mo regardless of where the server is. Your hardware or hosting cost is whatever you currently pay. There is no additional markup or per-server fee from OpenClawHosting. AI API usage is billed by the provider.

Your Server, Your Agent

Any Linux machine with SSH access. Total control, zero lock-in. Deploy now.