OpenClaw VPS requirements: how much RAM and CPU you need, and what it costs

By the serverserv team · 6 min read · Updated 15 October 2026

OpenClaw's official minimum is 1 GB of RAM, 1 CPU core, 500 MB of disk and a 64-bit OS, with 2 GB or more recommended. For daily personal use, plan on 2 vCPU and 4 GB. This guide covers the OpenClaw VPS requirements for each type of use, when you need more, whether a GPU helps, and what a month costs for the server and the AI model together.

What are the minimum OpenClaw VPS requirements?

OpenClaw officially needs 1 GB RAM, 1 CPU core, 500 MB of disk and a 64-bit operating system. The documentation recommends 2 GB of RAM or more for comfortable use.

The software is light because, with a cloud model, the server only runs the Gateway: a Node.js service that routes messages between your chat apps and your model provider. If you knew the project as Clawdbot or Moltbot earlier in 2026, it is the same software; it has been called OpenClaw since 30 January 2026.

• Operating system: Linux, macOS or Windows (WSL2 or native). Linux is the usual choice on a VPS.
• Runtime: Node 24.16+ or 26.1+, with Node 26 recommended. The installer adds Node if it is missing.
• Install: curl -fsSL https://openclaw.ai/install.sh | bash, then openclaw onboard --install-daemon to run it as a background service. Docker, Podman, Nix, Ansible and Kubernetes are also supported.
• Docker note: building the Docker image from source needs about 6 GB of RAM. Using the pre-built image avoids that.

OpenClaw's own DigitalOcean guide uses a 1 vCPU, 1 GB, 25 GB server with a 2 GB swapfile. That works for a test, but leaves little headroom.

How much RAM does OpenClaw need for real use?

For one person with a cloud model and a few chat channels, 4 GB of RAM is a comfortable size. Browser automation and local models are what push memory up.

Use	RAM	vCPU	Disk
Testing, one channel	1 to 2 GB	1	25 GB
Daily personal use, several channels	4 GB	2	50 GB
Browser automation (about 2 GB per headless browser)	8 GB	2 to 4	100 GB
Local model on the same server	16 GB or more	4 or more	200 GB

These bands match OpenClaw's documentation and third-party sizing guides. Vultr's guide, for example, suggests 2 vCPU and 2 GB for personal use and 4 vCPU and 8 GB for multi-channel use. If the server runs short of memory, the first sign is usually slow replies or a browser task that fails part way through.

How many CPU cores and how much disk does OpenClaw need?

One or two vCPU is enough for a personal agent on a cloud model, because the Gateway spends most of its time waiting for the model to reply. More cores help when the agent drives several browsers at once or runs a local model.

Disk use starts small. The app itself fits in 500 MB, but logs, skills, browser caches, Docker images and the agent's own files grow over time. 25 GB is a workable floor, and 50 GB gives comfortable headroom for normal personal use. Local models need far more, since a single model file can take several gigabytes.

A quick way to check your server after a week of use:

• free -h shows how much RAM is in use and whether swap is being touched.
• df -h shows free disk space.
• top shows which processes use the most CPU and memory.

If swap is busy most of the day, move up one plan size.

Does OpenClaw need a GPU?

No. With a hosted model, OpenClaw needs no GPU at all, because the model runs on the provider's systems. OpenClaw's Raspberry Pi guide puts it simply: the Pi "is just the gateway".

OpenClaw supports hosted providers including Anthropic, OpenAI, Google, OpenRouter and Amazon Bedrock. It also supports local runtimes: Ollama, LM Studio, vLLM and SGLang. A local model on a CPU-only server does work, but it needs 16 GB of RAM or more and replies far more slowly than a cloud API. Choose that route for privacy or to avoid API bills, not for speed. If fast local models are the main goal, a Mac may suit you better; see Mac mini vs VPS for OpenClaw.

What does an OpenClaw server cost per month?

An entry OpenClaw server with 2 to 4 GB of RAM costs about $12 to $16 a month at regular prices at the time of writing. Lower intro prices are common, but most renew higher.

Provider and plan	Specs	Intro price	Regular or renewal price
Hostinger KVM 1	1 vCPU, 4 GB, 50 GB	$6.49	$11.99
Bluehost NVMe 4	2 vCPU, 4 GB, 100 GB	$9.49	$11.99
IONOS VPS M+	2 vCPU, 4 GB, 120 GB	$5 for 3 months	$14
Elestio managed OpenClaw	managed	none	from $16
DigitalOcean 1-Click droplet	2 vCPU, 4 GB, 80 GB	none	$24
Vultr Marketplace (High Performance)	2 vCPU, 4 GB, 100 GB	none	$24

Prices are in US dollars per month, taken from each provider's pages or comparison articles at the time of writing, and they change often. Intro prices usually require a long prepaid term.

What does the AI model cost on top of the server?

The model is a separate bill from your model provider, and for most people it is the larger and less predictable part of the total. OpenClaw itself is free under the MIT licence and has no paid edition.

• API key, pay per use. Providers such as Anthropic, OpenAI, Google and OpenRouter charge by usage. A quiet personal agent costs little; an agent that reads long documents or browses all day costs much more.
• Subscription sign-in. OpenClaw can sign in with a ChatGPT or Codex subscription through OAuth, so a subscription you already pay for can cover the model.
• Local model. No API bill, but you need a server with 16 GB or more of RAM, and replies are slower.

So the monthly total is: server price + model usage. A quiet agent may cost less in model usage than the server itself, while an agent that browses or reads long documents all day can easily cost more. Set a monthly spending limit in your provider account before you connect busy channels. For full budgets by type of use, see how much OpenClaw costs per month.

Which serverserv plan meets OpenClaw's requirements?

OpenClaw Starter covers the recommended size for personal use with room to spare. Every OpenClaw plan comes with OpenClaw pre-installed and set up, and runs on CPU.

Plan	Specs	Monthly	3-year term	Good for
OpenClaw Starter	2 vCPU, 4 GB, 50 GB NVMe	$14.99	$10.99/mo	One person, cloud model
Personal Use	4 vCPU, 8 GB, 100 GB NVMe	$24.99	$17.99/mo	Browser automation
Power User	6 vCPU, 12 GB, 200 GB NVMe	$39.99	$28.49/mo	Several browsers, heavy use
Team Deployment	12 vCPU, 48 GB, 400 GB NVMe	$79.99	$57.49/mo	Small local models on CPU
Enterprise Scale	18 vCPU, 96 GB, 600 GB NVMe	$159.99	$114.99/mo	Larger local models on CPU

Prefer to install it yourself? A Linux KVM VPS such as KVM 1 (1 vCPU, 4 GB, 50 GB NVMe, $14.99 monthly) gives you full root access. Every plan renews at the price you paid, with at least 30 days' email notice if a price must change. See all plans on the OpenClaw hosting page, read running an AI agent on your own server, or open a ticket if a plan is not in the order form yet.

Comparing providers? See best OpenClaw hosting in 2026 for six options side by side, with renewal prices and 3-year totals.

Questions

Sources

• OpenClaw docs: Raspberry Pi install and hardware minimum
• OpenClaw on GitHub (install and runtime)
• OpenClaw docs: model providers
• Vultr docs: OpenClaw Marketplace application
• DigitalOcean Droplet pricing
• OpenClaw on Wikipedia (project history and names)

Prices from other companies are as published at the time of writing and can change.