English
Español
Português
Français
Română
Türkçe
Русский
فارسیOpenClaw VPS hosting
OpenClaw is an open-source AI assistant that works through your chat apps. On a VPS it runs around the clock instead of only while your laptop is on.
OpenClaw is an open-source AI agent you reach through Telegram, Discord, Slack, WhatsApp and other chat apps. It needs a machine that stays on, and Bulletproof.to gives it one, a KVM VPS in Vilnius or Tehran from €14.99 a month with full root access, paid in crypto and without ID.
What it needs
- Memory
- 2 GB RAM or more
- Disk
- 20 GB or more
- System
- Debian 13
- Smallest plan
- Iron, €14.99 / mo
The OpenClaw docs give no memory figure. Start with 2 GB for one agent that uses a hosted model and move to a bigger plan for browser automation or several agents.
Ports to open
| Port | Used for | Open to |
|---|---|---|
| 22/tcp | SSH, also for the tunnel to the dashboard | Only you |
| 18789/tcp | OpenClaw gateway and dashboard | Keep closed |
Set these in the free firewall in the panel, everything else stays closed.
Install OpenClaw
1. Install OpenClaw with the official script, it adds Node.js when it is missing
curl -fsSL https://openclaw.ai/install.sh | bash2. Run the setup, pick your model provider and chat apps, and install the background service
openclaw onboard --install-daemon3. Check that the gateway is running
openclaw doctor openclaw gateway status4. Open the dashboard from your own computer through an SSH tunnel, then visit http://127.0.0.1:18789
ssh -N -L 18789:127.0.0.1:18789 root@YOUR_SERVER_IP5. Run the security check before you connect real accounts
openclaw security audit
Commands from the official OpenClaw docs, checked in November 2026. The project may change them, so look there if one fails.
Why run OpenClaw here
On all the time
The agent keeps answering and running its jobs while your own computer is off.
Not tied to your name
No KYC and crypto payment, so the server your agent runs on is not linked to you.
Gateway stays private
The firewall keeps port 18789 closed to the internet, you reach the dashboard over SSH only.
Snapshots before updates
Take a snapshot in the panel before an upgrade and roll back in one click if something breaks.
Plans that fit
| Plan | vCPU | RAM | NVMe | Port | Per month |
|---|---|---|---|---|---|
| Iron | 1 | 2 GB | 20 GB | 1 Gbps | €14.99 |
| Steel | 2 | 6 GB | 40 GB | 1 Gbps | €22.99 |
| Cobalt | 4 | 8 GB | 60 GB | 2.5 Gbps | €29.99 |
| Chrome | 6 | 12 GB | 80 GB | 2.5 Gbps | €39.99 |
| Titanium | 8 | 16 GB | 100 GB | 5 Gbps | €49.99 |
| Tungsten | 12 | 20 GB | 120 GB | 5 Gbps | €59.99 |
| Carbide | 14 | 24 GB | 140 GB | 10 Gbps | €69.99 |
| Graphene | 16 | 32 GB | 160 GB | 10 Gbps | €79.99 |
Monthly prices. Pay for 3 months and save 5%, or 12 months and save 15%. Upgrade to a bigger plan from the panel any time.
Questions about hosting OpenClaw
How much RAM does OpenClaw need?
The official docs give no number. One agent that uses a hosted model such as Claude or GPT runs on 2 GB, browser automation and several agents need more.
Can I open the OpenClaw dashboard to the internet?
The docs advise against it. Keep port 18789 closed and reach the dashboard through an SSH tunnel or Tailscale.
Do I need a GPU?
No, not with a hosted model provider. Running a local model needs far more memory, which suits a dedicated server better.
Which chat apps can I use?
Telegram, Discord, Slack, WhatsApp and others, you pick them during setup.
Do you need my ID or name?
No. You sign up with an account number, no email or ID is required, and you pay with Monero, Bitcoin, USDT or Litecoin.
Do you install it for me?
No, you get a clean server with full root access and install it yourself with the steps on this page. Support helps with the server, not with the app.