We've got a new update. We now offer our own Web Console. Open the console and type ssh root@your-server-ip to reach any server you own, right from your browser. No SSH app needed.
Agents and automations need a machine that is always on, has its own IP and does not sleep when your laptop does. Start on a VPS from $8.50 a month for agent frameworks that call hosted models, or move to a GPU server from $109.25 a month to run the model yourself. Full root, NVMe storage, pay in crypto.
VPS prices are for Romania on a one-month term. A 12-month term takes 50% off. GPU prices are for Moldova.
or $4.25/mo on a 12-month term (−50%)
One or two agents or workflows that call hosted model APIs, plus a small database.
Configure and orderor $6.50/mo on a 12-month term (−50%)
n8n or a framework with Postgres, a vector store and a headless browser for web tasks.
Configure and orderor $12.75/mo on a 12-month term (−50%)
Small quantised models on CPU, slow but private, next to your agents.
Configure and orderor $54.63/mo on a 12-month term (−50%)
Run a model on the same server as the agent that uses it. Larger GPUs are on the GPU page.
Configure and ordern8n and similar tools: webhooks, schedules and integrations that run whether or not you are online.
LangGraph, CrewAI, AutoGen or your own Python and Node code, started by systemd and restarted if it fails.
Headless Chromium with Playwright for scraping, monitoring and form-filling tasks. Give it 8 GB of RAM.
Ollama or vLLM serving an open model, so prompts and documents stay on a server you rent.
Tools your assistant can call. See MCP server hosting for the details.
Telegram and Discord bots, trading bots and cron jobs that need to run all day.
Most agents do very little work themselves. They send a prompt to a model, wait, then call a tool. If the model runs at a provider, the agent server only needs a few gigabytes of RAM and a steady connection, so the starter VPS is enough.
An agent can run code and call tools, so treat it like any untrusted process. Give it its own Linux user or container, a model API key with a spending cap, and only the file and network access it needs. Never put a wallet's spending key or an admin password in an agent's environment.
Create an account, top up with crypto and deploy the size you picked. You get root access within minutes.
Update Ubuntu, create a user, open only the ports you need and install Docker.
Start n8n with Docker Compose, or run your own Python agent as a systemd service.
Put Caddy in front for HTTPS and install Ollama on a GPU server.
Commands assume Ubuntu 22.04 or 24.04 and a root login over SSH.
ssh root@YOUR_SERVER_IP
apt update && apt -y upgrade
adduser --disabled-password --gecos "" agent
ufw allow OpenSSH
ufw allow 80/tcp
ufw allow 443/tcp
ufw enablecurl -fsSL https://get.docker.com | sh
usermod -aG docker agent
systemctl enable --now dockern8n listens only on localhost here. Caddy (step 3) adds HTTPS in front of it. Replace agents.example.com with a domain that points at your server. Check the n8n hosting docs for the current options.
services:
n8n:
image: docker.n8n.io/n8nio/n8n
restart: unless-stopped
ports:
- "127.0.0.1:5678:5678"
environment:
- N8N_HOST=agents.example.com
- N8N_PROTOCOL=https
- WEBHOOK_URL=https://agents.example.com/
- GENERIC_TIMEZONE=UTC
volumes:
- n8n_data:/home/node/.n8n
volumes:
n8n_data:su - agent
cd ~/n8n
docker compose up -d
docker compose logs -f # Ctrl+C to stop watchingA systemd unit starts the agent at boot and restarts it if it crashes. Secrets live in a file only the agent user can read.
su - agent
mkdir -p ~/app && cd ~/app
python3 -m venv venv
./venv/bin/pip install requests # add your framework: langgraph, crewai, autogen...
nano agent.py # your agent code
nano .env # API_KEY=... (then: chmod 600 .env)
exit[Unit]
Description=My AI agent
After=network-online.target
Wants=network-online.target
[Service]
User=agent
WorkingDirectory=/home/agent/app
EnvironmentFile=/home/agent/app/.env
ExecStart=/home/agent/app/venv/bin/python agent.py
Restart=always
RestartSec=5
[Install]
WantedBy=multi-user.targetsystemctl daemon-reload
systemctl enable --now agent
journalctl -u agent -f # Ctrl+C to stop watchingCaddy gets and renews a certificate by itself once your domain points at the server. Webhooks and web interfaces then work over HTTPS.
apt -y install caddy
cat > /etc/caddy/Caddyfile <<'EOF'
agents.example.com {
reverse_proxy 127.0.0.1:5678
}
EOF
systemctl reload caddyOllama serves a model on localhost, and your agent calls it instead of a hosted API. Pick a model that fits the GPU memory (see the sizing table).
curl -fsSL https://ollama.com/install.sh | sh
ollama pull llama3.1:8b
curl http://localhost:11434/api/generate -d '{
"model": "llama3.1:8b",
"prompt": "Say hello in five words.",
"stream": false
}'# Ollama listens on 127.0.0.1:11434 by default. Leave it that way.
# If an agent on another server needs it, tunnel over SSH instead of opening the port:
ssh -N -L 11434:127.0.0.1:11434 agent@YOUR_SERVER_IPA server that runs your AI agents, automations and bots around the clock. You rent a VPS or GPU server with root access, install your agent software and leave it running.
Anything that runs on Linux: n8n, LangGraph, CrewAI, AutoGen, Flowise, Open WebUI, your own Python or Node code, and MCP servers. You have full root, so Docker works too.
Only to run the model yourself. If your agents call a hosted model API, a VPS is enough. A GPU server makes sense for local open models like Llama, Mistral or Qwen, or when prompts must not leave your server.
4 GB for a framework that calls hosted models, 8 GB with a database or a headless browser, and 32 GB or a GPU for local models. You can move to a bigger plan later.
Yes. Headless Chromium with Playwright works on a VPS. Plan about 500 MB to 1 GB of RAM per open page and give the server 8 GB if several run at once.
No. You create an account with a generated login, top up with crypto and deploy. See no-KYC VPS hosting. You stay responsible for what your agents do, under our terms.
From your account balance, topped up with Bitcoin, Monero, Ethereum, Litecoin, Solana, TRON, BNB, USDC or USDT. Longer terms take up to 50% off. See pricing.
Checked 20 October 2026. Chain sizes and software requirements change, so confirm against the source before you order.
Pick a size, pay in crypto, log in as root. No ID, no card.