🟢 All systems operational · Autonomous AI agents with persistent memory
+ Add Provider
Self-hosted Agent · Open source

OpenClaw — Your Personal
24/7 AI Agent

OpenClaw is an open-source personal AI agent you run yourself. It keeps working while you are away: it remembers context between sessions, reaches your terminal, files and a browser, and calls whichever model provider you plug into it. The machine it lives on is one you rent.

See what a box costs → Browse all VPS plans

Why OpenClaw

Persistent Memory
Your agent remembers context across sessions. Save facts, preferences, and project state for continuity.
Full Tool Access
Terminal, file system, web search, browser, image generation, and more — all at the agent's command.
Skill System
Teach your agent reusable workflows. Skills store procedures, API endpoints, and best practices.
Plugin Ecosystem
Extend capabilities with community plugins. Add custom tools and integrate with external services.
24/7 Automation
Schedule cron jobs, monitor systems, and run autonomous tasks while you sleep.
Multi-Platform
Deploy on your own infrastructure or use managed hosting. Full control over your agent's environment.

What the box actually has to do

These four steps are ours, not OpenClaw’s. The project publishes no hardware minimum, and we are not going to invent one. What its own README does say shapes the answer: it installs with a single command, wants a current Node.js, runs on Linux, macOS or Windows with WSL2, opens a browser-based Control UI, and plugs into swappable model providers — hosted ones included. Point it at a hosted model and no GPU is involved, so this is a small always-on box question. The browser is the part that decides the size: an agent that only chats is far lighter than one driving a page.
Setup What is running Our sizing step
Trying it out The agent process and its Control UI, talking to a hosted model 1 core / 2 GB
Daily driver The same, left running, with memory, skills and plugins accumulating on disk 2 cores / 4 GB
Browser and media work A headless browser driven as a tool alongside the agent, plus scheduled jobs 4 cores / 8 GB
Heavy automation Several tasks at once, longer histories, bigger artefacts kept around 8 cores / 16 GB

Two costs this does not cover. The model bill is separate and goes to whichever provider you point the agent at, not to the host. And if you decide to serve that model yourself instead, the machine question changes completely: that becomes a GPU with enough memory to hold the weights, which is what the DeepSeek page and the GPU catalogue are for.

Cheapest server at each step

All VPS plans →

For every rung above, the cheapest VPS in the DohoHub catalogue that meets it right now, whoever sells it. These are the prices our feed already tracks, normalised so they can be compared and shown in the currency picked in the header. When one machine is the cheapest way to clear two rungs, it appears once and the note says which rungs it covers.

Provider Plan Specs Price Visit
Database Mart
Express Linux VPS
cheapest from 1 core / 2 GB RAM up to 2 cores / 4 GB RAM
 2 vCPU4 GB RAM60 GB SSD
$2.88/mo
 Visit
Contabo
Cloud VPS 12 (12 GB RAM)
cheapest at 4 cores / 8 GB RAM
 6 vCPU12 GB RAM100 GB NVME0.2 Gbps
$3.92/mo
renews at $5.22
 Visit
Contabo
Cloud VPS 24
cheapest at 8 cores / 16 GB RAM
 8 vCPU24 GB RAM200 GB NVME0.3 Gbps
$6.09/mo
renews at $8.12
 Visit

Ordinary VPS plans, not an OpenClaw product — the install command is yours to run. An agent you talk to all day wants to stay up, so read the provider’s uptime record on its page and not only the price. Where a headline rate only buys an introductory term, the renewal rate is printed beneath it.

Named by the provider

Contabo on DohoHub →

One company in the catalogue goes further and markets specific machines for self-hosted OpenClaw: Contabo lists four of its Cloud VPS servers by how much work you expect the agent to carry. They are its ordinary Cloud VPS machines — same hardware, same price as everywhere else in our catalogue — so the figures below come from the same price feed. Nobody paid for this section; it is here because the claim is theirs and no other provider makes one.

Provider Plan Specs Price Visit
Contabo
Cloud VPS 4
OpenClaw Personal Use
 4 vCPU8 GB RAM100 GB SSDUnlimited traffic0.2 Gbps
$5.10/mo
renews at $6.38
 Visit
Contabo
Cloud VPS 6
OpenClaw Power User
 6 vCPU12 GB RAM200 GB SSDUnlimited traffic0.3 Gbps
$6.96/mo
renews at $8.70
 Visit
Contabo
Cloud VPS 12
OpenClaw Team Deployment
 12 vCPU48 GB RAM400 GB SSDUnlimited traffic0.8 Gbps
$23.20/mo
renews at $29.00
 Visit
Contabo
Cloud VPS 18
OpenClaw Enterprise Scale
 18 vCPU96 GB RAM600 GB SSDUnlimited traffic1 Gbps
$45.47/mo
renews at $56.84
 Visit

Contabo’s headline rate buys the first 24 months; the renewal rate is printed beneath it. Prices are published by Contabo including VAT. We hold them in euro and convert at the market rate — but Contabo also sets its own dollar and pound prices, and those are not that conversion, so the amount it charges you outside the euro area can be noticeably higher than the figure above. Check the total on Contabo’s own checkout before you buy.

Managed OpenClaw

The other way to run it: the provider hosts and updates OpenClaw for you. There are no server specs to compare here — these are managed services, not machines you rent, so the list below shows only what each provider states is included.

Managed OpenClaw
by Hostinger
Managed for youReady out of the boxNo maintenance neededVisual interface includedCLI access included
$5.99/mo
renews at $11.99
Visit

Managed plans are billed on the provider’s own promotional terms — the renewal rate is printed beneath each price. Hostinger publishes no CPU, RAM or storage figures for this product and describes it as an alternative to running OpenClaw on a VPS, so none are shown.

Rent the box, then run one command

Filter the whole catalogue by cores, memory, storage and location — live prices, tracked every six hours.

Find a server by specs →
↑