OpenClaw · Always on · Private gateway

OpenClaw VPS Your assistant stays online with no Mac mini at home

A Linux VPS for running the OpenClaw Gateway around the clock. It sits in a datacenter with a fixed IP, apart from your personal computer, and you reach it through an SSH tunnel. Full root, reserved memory and monthly billing. The 4 GB plan we suggest starts at $14.99₹1,389.27 a month.

OpenClaw VPS plans

OpenClaw publishes no memory minimum, so these sizes are our suggestion: 2 GB for a light chat agent, 4 GB when the agent also drives a headless browser, 8 GB for several agents. Billed monthly, with no long contract.

Compute Region

Choose the deployment location
Currently out of stock

OpenClaw VPS plans · Miami

Miami, Florida · AMD EPYC KVM VPS with full root access and a choice of 15 Linux builds
Plan vCPU Memory Storage Bandwidth Setup Price
4GB Plan Recommended 4 vCPU 4 GB 80 GB SSD 5 TB Instant $14.99/mo₹1,389.27/mo Buy Now
6GB Plan 4 vCPU 6 GB 100 GB SSD 5 TB Instant $19.99/mo₹1,852.67/mo Buy Now
8GB Plan 6 vCPU 8 GB 100 GB SSD 5 TB Instant $29.99/mo₹2,779.47/mo Buy Now
16GB Plan 8 vCPU 16 GB 100 GB SSD 5 TB Instant $59.99/mo₹5,559.87/mo Buy Now
Choose Ubuntu 24.04 on the order form. The OpenClaw installer adds the Node 24 it needs.

Why run OpenClaw on a server and not at home

OpenClaw is an assistant you message from your chats. It can only answer while the machine behind it is awake.

It answers at any hour

The gateway runs as a service on a machine that does not sleep, travel or restart for a system update halfway through a task.

A gateway nobody else can reach

OpenClaw keeps its gateway on loopback, port 18789. With the firewall allowing SSH only, the one way in is your own SSH tunnel.

Away from your own files

An assistant that can run commands is safer on a machine that holds nothing personal. The server has only what you put on it.

One fixed address

Each server has its own IPv4 address that stays the same. Your tunnel, your SSH shortcut and any service that needs to know where your agent calls from keep working.

Memory that is not shared out

The RAM in your plan is reserved for your server. Running short of memory is a common reason a self-hosted agent falls over, so this matters.

A clean start when you want one

If a setup goes wrong, reinstall the operating system from the client area and run the six steps again. It is quicker than untangling a broken install.

OpenClaw VPS setup in 6 steps

Every OpenClaw command below comes from OpenClaw's own documentation. The rest is ordinary Linux housekeeping.

  1. 1

    Order a plan with Ubuntu 24.04

    Click Buy Now on a plan above and choose Ubuntu 24.04 on the order form. The server is built automatically after payment, and its IP address and root password arrive by email.

  2. 2

    Create a normal user and log in as that user

    Do not install OpenClaw as root. Log in as root once, create a user, log out, and come back in as the new user. Replace YOUR_SERVER_IP with the address from the email.

    ssh root@YOUR_SERVER_IP
    adduser claw
    usermod -aG sudo claw
    exit
    ssh claw@YOUR_SERVER_IP

    Log in directly, not with su. The gateway is installed as a service of this user, and Linux starts a user's services only for a real login.

  3. 3

    Turn on the firewall with only SSH open

    Allow SSH first, then switch the firewall on. The gateway port is not opened, and it does not need to be.

    sudo apt update && sudo apt install -y ufw
    sudo ufw allow 22/tcp
    sudo ufw enable
  4. 4

    Run the official installer

    OpenClaw needs Node 24.16.0 or newer on the 24.x line, or 26.1.0 or newer, as of November 2026. You do not have to install it by hand: on Linux the installer adds Node 24 LTS for you.

    curl -fsSL https://openclaw.ai/install.sh | bash
  5. 5

    Onboard and install the gateway service

    Onboarding asks which model to use. It can reuse an existing Claude Code or Codex CLI login, or take a provider API key. The --install-daemon flag sets the gateway up as a background service.

    openclaw onboard --install-daemon
    openclaw gateway status
    sudo loginctl enable-linger $USER

    The last line is a Linux setting, not an OpenClaw one. It lets this user's services keep running after you log out. To manage the service by hand, use openclaw gateway install, openclaw gateway start and openclaw gateway status.

  6. 6

    Reach the gateway through an SSH tunnel

    The gateway listens on loopback, port 18789, so only the server itself can talk to it. Run this on your own computer, with your user and server IP in place of user@gateway-host.

    ssh -N -L 18789:127.0.0.1:18789 user@gateway-host

    While the command runs, port 18789 on your own computer leads to the gateway on the server. OpenClaw's VPS guide names Tailscale Serve as the other supported way in.

Why OpenClaw on a VPS can be painful, and how to avoid it

Search for OpenClaw on a VPS and you will find people who call it painful. The trouble usually comes from one of three things, and each has a plain fix.

If you do change the bind to lan or tailnet, OpenClaw requires gateway.auth.token or gateway.auth.password to be set.
What goes wrongWhy it happensWhat the steps above do about it
Too little RAMThe agent, Node and perhaps a headless browser all need memory. On a server that is too small the system kills whichever process is largest, often in the middle of a task.Start at 2 GB for chat only and 4 GB with a browser. The memory in the plan is reserved for you.
Running as rootEverything the agent does then has full control of the server, and files end up owned by root, which breaks later commands run as a normal user.Step 2 creates a normal user before anything is installed, and the gateway runs as a service of that user.
Exposing the gatewayChanging the gateway to listen on the public address so that it is easy to open from a laptop puts a tool that can run commands on the open internet.The gateway stays on loopback, the firewall allows SSH only, and you connect through the tunnel in step 6.
What a VPS will not fix. A server does not remove the learning curve. You still work in a terminal, and you still pay your model provider separately. If you want a managed service where somebody else runs OpenClaw for you, this is not that. We give you the machine and help with the machine.

OpenClaw on a Mac mini vs a VPS

One is bought once and sits at home. The other is rented by the month and sits in a datacenter with a fixed IP.

Running OpenClaw on your own laptop works too, for trying it out. It stops answering when the lid closes.
Mac mini at homeVPSLake VPS
How you payOnce, and the machine is yoursBy the month, from $14.99₹1,389.27 for 2 GB
Where it sitsAt home, on your home internetIn a datacenter
Same public IP every dayUsually not on a home line
Keeps running through a home power or internet cut
On the same network as your personal devicesYesNo
Can use things that need macOS itselfNo. It is a Linux server
Can run a local modelSmaller ones, depending on its memoryNo. There is no GPU
Start again from a clean systemReinstall macOS by handReinstall from the client area

Which plan for which OpenClaw setup

OpenClaw states no official RAM minimum as of November 2026. These sizes are our suggestion, not a vendor requirement.
What you runPlan we suggestWhy
One light chat agent that calls a model over the internet2 GBThe server carries only the gateway and Node. The model does its work somewhere else.
An agent that also drives a headless browser4 GBA browser is the hungriest thing on the server. 4 GB leaves room for it next to the gateway.
Several agents at once8 GBEach agent and each browser takes its own share of memory.
A large local model on the server itselfNot these serversThere is no GPU. That job belongs on a GPU machine.

What an OpenClaw VPS is

OpenClaw is a self-hosted assistant gateway from the OpenClaw Foundation. Its own tagline is "Your assistant, on your devices, in your chats". You may know it by its earlier names, Clawdbot and Moltbot. An OpenClaw VPS is a Linux server in a datacenter that runs that gateway for you all day. Every VPSLake plan is a private KVM virtual machine with full root, reserved memory, SSD storage and its own IPv4 address.

It suits people who want the assistant reachable at any hour, and who would rather not give an agent the run of the computer that holds their own files. If you only want to try OpenClaw for an afternoon, install it on the computer you already have. You do not need a server for that.

What we do and do not set up for you

Our part is the server: Ubuntu freshly installed, root access and one fixed IP address. Your part is OpenClaw. It is not preinstalled and we have no one-click installer, so you paste the project's own install line and land on its current release. The model behind your assistant is billed by whoever provides it, never by us. When the server itself misbehaves, for example SSH will not connect or you want a reinstall, write to support by ticket, WhatsApp or Telegram on any day.

The model is not on the server

These servers have no GPU, so OpenClaw here calls a model over the internet. Onboarding can reuse an existing Claude Code or Codex CLI login, or a provider API key. If you want to reuse a CLI login, install and sign in to that tool on the server first. Our Claude Code VPS and Codex VPS pages show how to sign in to each on a server with no browser.

Linux, not Windows

We also sell Windows remote desktops, but they are the wrong product for this. OpenClaw's documentation says that for the native Gateway "Windows Server is not supported". Order a Linux VPS from this page.

Where the servers are

You choose the location in the plan table: Miami, Florida, Frankfurt, Germany or Amsterdam, The Netherlands. For OpenClaw the location matters little, because the gateway only exchanges text with model APIs and chat services. Comparing agents? Hermes Agent VPS covers a similar always-on agent, and AI agent VPS lists the common ones side by side.

OpenClaw is a product of the OpenClaw Foundation. VPSLake is an independent hosting provider and is not affiliated with the OpenClaw Foundation.

Pay the way you already pay

  • USDT, BTC & 20+ coinsvia Cryptomus
  • UPI / PhonePe / PaytmINR via Razorpay
  • Visa / Mastercardvia Paddle
  • Net banking & walletsIndia

Frequently Asked Questions

Yes. OpenClaw's own documentation has a page for Linux servers, and its installer adds the Node 24 it needs on Linux. You connect by SSH, run the installer as a normal user, and the gateway runs as a background service. The six steps on this page are the whole job.
One with enough memory, full root and a fixed IP. OpenClaw gives no official RAM minimum, so the sizes here are our suggestion: 2 GB for a light chat agent, 4 GB when the agent also drives a headless browser, and 8 GB for several agents. If you are unsure, take 4 GB. Moving up later keeps your data and needs only a reboot.
Our suggestion for the smallest sensible size is the 2 GB plan, for one light chat agent that calls a model over the internet. Its current price is in the plan table. We have no free plan and no free trial, but billing is monthly, so you can try it for one month and stop.
No, and there is no one-click installer. You get a clean Ubuntu or Debian server with root access and you paste OpenClaw's official installer command yourself. That way you run the version OpenClaw ships today, not an old image somebody prepared.
No. The VPS is the computer OpenClaw runs on. The model comes from your own account: onboarding can reuse an existing Claude Code or Codex CLI login, or take an API key from a model provider. We do not sell subscriptions or API credits.
A Mac mini is bought once and sits at home on your home internet. A VPS is rented by the month and sits in a datacenter with a fixed IP, away from your personal devices. Choose the Mac if you need macOS itself or want to try local models. Choose the VPS if you want the agent off your home network and reachable every hour of the day.
Do not open it. The gateway binds to loopback on port 18789, which means only the server itself can reach it, and OpenClaw's VPS guide says to keep it that way and connect through an SSH tunnel or Tailscale Serve. The firewall step on this page allows SSH only.
An API key from your model provider is the simplest route, because it needs no browser. If you would rather use a Claude or ChatGPT plan, sign in to the Claude Code or Codex CLI on the server first, and OpenClaw onboarding can reuse that login. Claude Code prints a link you open on your own device, and Codex has codex login --device-auth.
Yes. A VPS is a virtual machine that you rent from a host, with its own operating system, memory and IP address. Ours are private KVM virtual machines, so you get full root and can run Docker. To OpenClaw it looks like any other Linux computer.
Yes. Switch the price to INR with the currency button in the menu and pay by UPI. Crypto and card work too. The servers are in the USA and Europe, which is fine for OpenClaw, because the gateway talks to model APIs and chat services over the internet.

Put OpenClaw on a machine that stays awake

Order a VPS, run the official installer and keep the gateway private behind an SSH tunnel. Billed monthly, cancel when you like.

Related solutions