Best Cheap VPS for OpenClaw 2026: $5 vs $10 vs $20/mo Compared

Updated March 2026 · 7 min read

OpenClaw is an open-source AI agent framework that runs 24/7 on your own server. It needs real resources — 2 vCPU, 4 GB RAM, and 40 GB SSD at minimum. If you want browser automation, bump that to 8 GB RAM. The cheapest VPS plans on the market start at $2/mo, but most of those won't run OpenClaw at all. We dug through our database of 40+ providers to find the cheapest VPS plans that actually work.

This guide is organized by budget. Jump to the tier that matters to you, or read straight through — we'll end with a summary table showing exactly what we'd buy at each price point.

Can You Run OpenClaw on a $5 VPS?

Barely — and in most cases, no. Here's the problem: the cheapest VPS plans on the market give you 512 MB to 2 GB of RAM. OpenClaw's official minimum is 4 GB RAM. That's not marketing padding; it's the Docker container running Node.js 22+, loading your agent config, maintaining WebSocket connections to Telegram/Discord/Slack, and keeping conversation context in memory.

The ultra-cheap providers in our database look tempting:

• IONOS at $2/mo — 1 vCPU, 512 MB RAM. Not even close.
• Vultr at $2.50/mo — 1 vCPU, 512 MB RAM. Same story.
• UpCloud at $3/mo — 1 vCPU, 1 GB RAM. OpenClaw will start but OOM-kill within hours.
• Hetzner at $3.49/mo — 2 vCPU, 4 GB RAM (EU only). This one actually meets the minimum.
• BuyVM at $3.50/mo — 1 vCPU, 1 GB RAM. Text-only mode might limp along, but expect instability.

Can you technically get OpenClaw to start on a 1 GB VPS? Yes, if you disable all browser-based skills, strip the config to a single text-only agent, and add 2 GB of swap. But you'll hit constant memory pressure, slow responses, and random crashes. We tested this on an UpCloud $3 instance and the agent survived about 4 hours before the kernel OOM-killer stepped in.

The exception is Hetzner's CX22 at $3.49/mo, which gives you 2 vCPU and 4 GB RAM — exactly the OpenClaw minimum. The catch: it's EU-only (Falkenstein/Helsinki/Nuremberg), so latency from North America is 100–150 ms. If your agent doesn't need low-latency API calls, this is the cheapest way to run OpenClaw properly. It's technically under $5 and it works.

Verdict: A $5 VPS can run OpenClaw only if that $5 gets you 4 GB+ RAM. Hetzner at $3.49 is the standout. Everything else at this price point gives you 1–2 GB, which is below the safe minimum.

Best VPS Under $10/mo for OpenClaw

This is the sweet spot. Once you cross the $5 line, several providers offer plans that comfortably meet or exceed OpenClaw's requirements. Here's what we found:

Provider	Price	vCPU	RAM	Storage	OpenClaw Ready?
Contabo	$4.95/mo	4 vCPU	8 GB	75 GB NVMe	Yes — exceeds minimum
Hetzner	$7.11/mo	2 vCPU	4 GB	40 GB SSD	Yes — meets minimum
Hostinger	$6.49/mo	1 vCPU	4 GB	50 GB NVMe	Yes — has Docker template
DigitalOcean	$6/mo	1 vCPU	1 GB	25 GB SSD	No — not enough RAM

Contabo at $4.95/mo is the clear winner. You get 4 vCPU, 8 GB RAM, and 75 GB NVMe storage. That's not just enough for OpenClaw — it's enough for OpenClaw plus a PostgreSQL database, Redis cache, and still have headroom. No other provider comes close at this price. The trade-off is that Contabo's network speeds are mediocre (200 Mbit/s) and their support is slow, but for a self-hosted AI agent that mostly makes outbound API calls, neither of those matters much.

Hostinger at $6.49/mo deserves a mention because they offer an OpenClaw Docker template in their VPS control panel. Instead of SSHing in and running docker compose up yourself, you pick OpenClaw from a dropdown and it's deployed in minutes. The specs (1 vCPU, 4 GB RAM) are tight but workable for a single agent. If you're not comfortable with the command line, Hostinger's template saves you real setup time.

DigitalOcean's $6/mo droplet is a popular recommendation in Reddit threads, but it only gives you 1 GB RAM — far below what OpenClaw needs. Their $4/mo plan is even worse at 512 MB. You'd need to jump to DigitalOcean's $24/mo tier to get 4 GB RAM, which is 5x the cost of Contabo for the same specs.

To deploy on any of these, SSH into your VPS and run:

git clone https://github.com/openclaw/openclaw.git
cd openclaw
cp .env.example .env
# Edit .env with your API key (Anthropic, OpenAI, etc.)
docker compose up -d

For the full deployment walkthrough, see our Docker 1-Click Deploy guide.

Best VPS $10–20/mo for Browser Automation

OpenClaw's browser automation skills (web scraping, form filling, screenshot capture) use a headless Chromium instance via Puppeteer or Playwright. Each browser tab consumes 200–500 MB of RAM on top of OpenClaw's base footprint. If you're running 2–3 concurrent browser tabs, you need 8 GB RAM minimum.

Here's the paradox: Contabo already gives you 8 GB at $4.95/mo. So if your only requirement is browser automation on a budget, Contabo is still the answer, and you don't need to spend $10–20 at all.

But there are legitimate reasons to spend more:

• Better CPU performance. Contabo's 4 vCPUs are shared and can be throttled under load. Premium providers give you dedicated or burstable cores that handle Chromium rendering faster.
• US/Asia data centers. Contabo's cheapest plans are EU-only. If you need a US or Asia location, you'll pay more.
• Managed support. If something breaks at 3 AM, DigitalOcean's support is faster than Contabo's.

In the $10–20 range, here's what makes sense for browser-heavy OpenClaw workloads:

Provider	Price	vCPU	RAM	Storage	Notes
Contabo	$4.95/mo	4	8 GB	75 GB NVMe	Best value; EU location
InterServer	$12/mo	2	8 GB	160 GB SSD	US data center; price-lock guarantee
Hostinger	$13.49/mo	2 vCPU	8 GB	100 GB NVMe	Docker template; good UI
DigitalOcean	$24/mo	2	4 GB	80 GB SSD	Premium network; US/EU/Asia

Notice that DigitalOcean's $24/mo plan only gives you 4 GB RAM — half of what Contabo offers at $4.95. You're paying a 5x premium for DigitalOcean's faster network, better support, and global data center presence. Whether that's worth it depends on your use case. For most people self-hosting OpenClaw as a personal agent, it's not.

If you need browser automation in the US and don't want to deal with Contabo's EU latency, InterServer at $12/mo is a solid middle ground. They're based in New Jersey, offer a price-lock guarantee (your rate never increases), and 8 GB RAM handles Chromium comfortably.

The Free Option: Oracle Cloud Always Free

Oracle Cloud's Always Free tier is the best-kept secret in VPS hosting. You get 4 ARM Ampere CPU cores, 24 GB RAM, and 200 GB storage — permanently free, not a trial. That's more RAM than most $40/mo plans from other providers.

OpenClaw runs beautifully on this hardware. You can run multiple agents, a PostgreSQL database, Redis, browser automation with 10+ tabs, and still have RAM to spare. We've seen production OpenClaw deployments on Oracle's free tier that have been running continuously for months.

The catches:

• ARM architecture (aarch64). Most OpenClaw Docker images are built for x86_64. You'll need to use multi-arch images or build from source with docker buildx. The OpenClaw team ships ARM images, but some community plugins don't.
• Waitlist. Oracle's free tier is so popular that new account signups are frequently waitlisted, sometimes for weeks. Availability depends on the region you select — US-East (Ashburn) and EU-Frankfurt tend to be the hardest to get.
• Account termination risk. Oracle has been known to reclaim idle Always Free instances. Keep your agent active and make sure it's generating some compute usage to avoid being flagged.
• Egress limits. You get 10 TB/mo of outbound bandwidth, which is generous, but not unlimited.

To deploy OpenClaw on Oracle's ARM instance:

# SSH into your Oracle Cloud instance
sudo apt update && sudo apt install -y docker.io docker-compose-plugin
git clone https://github.com/openclaw/openclaw.git
cd openclaw
# Use ARM-compatible image
docker compose up -d

Verdict: If you can get an account, Oracle Cloud's free tier is unbeatable. But don't rely on it as your only option — the waitlist is real, ARM compatibility issues exist, and Oracle could change the program's terms. Treat it as a bonus, not a strategy.

What We'd Pick at Each Budget

Here's our recommendation for every price point, based on the plans currently in our OpenClaw VPS comparison database:

Budget	Provider	Specs	Verdict
$0	Oracle Cloud Free	4 ARM / 24 GB / 200 GB	Best specs on paper; waitlist & ARM caveats
$3–5	Hetzner CX22	2 vCPU / 4 GB / 40 GB	Cheapest x86 plan that meets minimum; EU only
$5–10	Contabo Cloud VPS 10	4 vCPU / 8 GB / 75 GB NVMe	Best value overall — 2x minimum specs
$5–10	Hostinger KVM 1	1 vCPU / 4 GB / 50 GB NVMe	Best for beginners (Docker template)
$10–20	InterServer 4 Slice	2 vCPU / 8 GB / 160 GB SSD	Best US option for browser automation

If you're reading this and just want one answer: Contabo at $4.95/mo. It gives you double the minimum RAM, quadruple the CPU, and enough storage for months of conversation logs. It's not the fastest VPS on the market, but for running an AI agent that makes outbound API calls, speed doesn't matter nearly as much as headroom.

For a full, filterable comparison of every VPS plan that meets OpenClaw's requirements, see our OpenClaw VPS comparison tool.

Frequently Asked Questions

What are the minimum VPS specs for OpenClaw?

The official minimum is 2 vCPU, 4 GB RAM, and 40 GB SSD. You also need Node.js 22+ and Docker (recommended). For browser automation skills, you'll need 8 GB RAM to handle headless Chromium alongside the agent process. Plans with 1–2 GB RAM will crash under normal use.

Is Contabo good enough for OpenClaw, or is it too slow?

Contabo is excellent for OpenClaw. The AI agent pattern — receiving a message, calling an LLM API, returning a response — is not CPU- or network-intensive on the VPS side. The bottleneck is the external API call, not your server's raw performance. Contabo's 4 vCPU / 8 GB at $4.95/mo gives you more headroom than a DigitalOcean droplet at 5x the price.

Can I run OpenClaw on Oracle Cloud's free tier?

Yes. Oracle's Always Free tier provides 4 ARM CPU cores and 24 GB RAM, which is more than enough. The main challenges are ARM architecture compatibility (use multi-arch Docker images or build from source), the waitlist for new accounts, and the risk of Oracle reclaiming idle instances. If you get an account, it's the best free option available.

Should I pick a VPS close to my LLM API provider's servers?

It helps, but it's not critical. The latency between your VPS and the LLM API (Anthropic, OpenAI) adds 10–50 ms to each request, while the API itself takes 1–10 seconds to generate a response. Choosing a US-based VPS when using US-hosted APIs saves you maybe 100 ms per request — barely noticeable. Pick your VPS based on price and specs first, location second.