VPS Pro 1
1 vCPU · 2 GB RAM · 20 GB NVMe
For a personal agent, a bot or a handful of automations.
CA$6.99/month
Annual billing: CA$76.89, one month free
Managed version: +CA$21 per month
AI VPS
AI VPS: the server that keeps your AI agents running day and night
Simafri's AI VPS keeps your agents, your automations and your AI applications online around the clock, on Intel Xeon processors and NVMe SSDs. Designed by an artificial intelligence expert who runs his own agents on it, it includes Full-Stack Assistance: every question gets a diagnosis and precise guidance.
- Online 24/7
- Webhooks over HTTPS
- Full-Stack Assistance included
Designed by an AI expert
A VPS designed by someone who runs his own agents on it
My name is Simon Vasconcelos Lee. I have been building websites since 2002, and I have been studying artificial intelligence systems since TensorFlow was released in 2015.
My own agents run day and night on our servers, as does the AI assistant I develop with every day. So I designed the AI VPS around what those agents truly need: staying online without interruption, keeping their API keys out of harm's way, receiving their webhooks over HTTPS and finding their memory intact after every restart.
My own agents call for a large configuration. Most projects get off to a very good start on one of the three sizes below, and the server then grows along with them.
Simon Vasconcelos Lee's personal websiteUse cases
What you can run on an AI VPS
An AI agent keeps working while you sleep, so it needs a server that never switches off. Here is what our clients set up on theirs.
A personal agent
OpenClaw and agents like it answer on WhatsApp, Telegram or Slack, sort your email and launch tasks on your behalf. On the AI VPS, they stay reachable even when your computer is off.
n8n automations
Your workflows call an AI model for every form submitted, every invoice that lands, every incoming message, triggered by a webhook or on a schedule.
A messaging bot
An assistant answers your customers on Telegram, Discord or WhatsApp Business at any hour, in your company's voice and with its information.
An app that calls an AI API
Your Next.js, FastAPI or Laravel application queries models from OpenAI, Anthropic, Mistral or Google, while its users, its database and its files stay with you.
Search across your documents
PostgreSQL and pgvector store your documents as vectors: your assistant finds the right passage before it answers, drawing on your own data.
A compact model on the server
Ollama or llama.cpp run embedding, classification or summarization models on the CPU, for background processing at a fixed cost.
Pricing
Three sizes, from your first agent to an app in production
An AI VPS is a VPS Pro shaped for AI: root access, a nightly backup copy, constant monitoring and Full-Stack Assistance. Pick the size that matches what you run; when your project grows, your server steps up a size and keeps your data.
VPS Pro 2
2 vCPU · 4 GB RAM · 50 GB NVMe
For an agent that drives a browser, n8n and its workflows, a first AI application.
CA$12.49/month
Annual billing: CA$137.39, one month free
Managed version: +CA$35 per month
VPS Pro 3
2 vCPU · 6 GB RAM · 80 GB NVMe
For an AI application with PostgreSQL and pgvector, or a compact model on the server.
CA$20.99/month
Annual billing: CA$230.89, one month free
Managed version: +CA$49 per month
Monthly prices excl. tax. Running a model with 7 billion parameters or more on the server? See sizes Pro 4 to Pro 6 on the VPS Pro page.
Built for AI
What an agent needs, ready from day one
Always online
We keep watch over your server without letup: your agent stays reachable at night, on weekends and while you are on vacation.
Protected API keys
Your keys live on a server where you are the sole administrator, reached over SSH on a port of its own, behind permanent anti-DDoS filtering.
Webhooks over HTTPS
WhatsApp, Telegram, Stripe or GitHub reach your agent at an HTTPS address, and we issue and renew its certificate.
Scheduled tasks
A summary every morning, a follow-up every evening: cron and systemd start your agents right on time, on a server that never sleeps.
Your stack, installed as root
Docker, Node.js, Python, PostgreSQL: you install whatever your agent asks for, in the exact version it asks for.
Its memory, backed up
Conversation history, vector database, configuration: each night, the entire server is encrypted and copied away from the machine, and our team brings it back whenever you ask.
Full-Stack Assistance
A question about your agent? We diagnose it
Included with every AI VPS, Full-Stack Assistance begins with what you notice, on the server or in your agent. We track down the cause, then send you the way forward, which you apply yourself or pass on to your AI assistant.
My agent stops as soon as I close my SSH session.
We show you how to run it as a service, with systemd or Docker, so that it starts again on its own after an outage or an update.
WhatsApp webhooks never reach my automation.
We check the public address, the certificate and the proxy with you, then give you the setting that is missing.
My server has been short of memory since I added vector search.
We measure how much room the vectors and their index take up, then point you to the setting or the size that fits.
How do I protect my API keys?
We show you where to keep them, how to restrict who can read them and which ports to close on your server.
Would you rather not touch the server at all? With Managed VPS, we take the server off your hands and keep a proactive eye on your agent. Explore Managed VPS
Tech specs
The hardware beneath your agent
AI VPS plans run on the same machines as our production servers, chosen to work day and night.
- Processor
- Server-class Intel Xeon, 1 to 2 vCPU depending on the size
- Memory
- 2 to 6 GB of ECC RAM, protected against bit errors
- Storage
- 20 to 80 GB of datacenter NVMe SSD, in a mirror
- Access
- Root, with SSH on a port reserved for your server
- Operating systems
- Debian, Ubuntu, Rocky Linux or AlmaLinux
- Network
- A public IPv6 address of your own, anti-DDoS always on
- Backup
- The whole server copied and encrypted every night, one week of history off the machine
- Growth
- Step up to Pro 4, 5 or 6 without reinstalling, data in place
VPS for AI: calling an API or hosting the model?
A VPS for AI mostly runs code that calls a model. In the vast majority of projects, the model is hosted by its provider and reached through an API: your server prepares the request, waits for the reply and acts on it. The processor has little to do, which is why a Pro 1 or a Pro 2 already carries a complete agent.
Hosting the model yourself keeps your data at home and holds the cost steady, whatever the number of requests. On a CPU, compact models are a great fit: embeddings, message classification, document summaries as background jobs. To size a specific project, our guide what VPS you need to run AI lists the memory footprint of common models.
Why a server rather than your own computer
An agent installed on a laptop stops when the lid closes and misses its webhooks the moment the Wi-Fi drops. On an AI VPS, it has a public address, an HTTPS certificate, a nightly backup and constant monitoring: it works for you at any hour, and you keep full control over everything it does.
Frequently asked questions
Your questions about the AI VPS
What is an AI VPS?
It is a private server, switched on at all times, where your artificial intelligence agents, automations and applications run. At Simafri, the AI VPS takes three sizes of the VPS Pro, with Intel Xeon processors, ECC memory, NVMe drives and Full-Stack Assistance.
Do I need a graphics card to run an AI agent?
Not for an agent that calls a model through an API: the computing happens at the model provider, and your server orchestrates. The AI VPS processor also handles compact models: embeddings, classification, summaries as background jobs.
Can I install OpenClaw on an AI VPS?
Yes. With root access, you install it just as you would on your computer, and it stays connected to your messaging apps day and night. We recommend Pro 2 as soon as your agent drives a browser, which is hungry for memory.
Can I host n8n on it?
Yes, with Docker or Node.js. Your workflows receive their webhooks over HTTPS and call the AI models of your choice. Plan on Pro 2 for n8n and its PostgreSQL database.
Are my API keys safe?
They stay on your server, and you are its only administrator. You connect over SSH on a port unique to your VPS, anti-DDoS filtering never switches off, and Full-Stack Assistance shows you where to store your keys and which ports to close.
Can I run a model with Ollama?
Yes, for compact models. A model with one to three billion parameters, quantized to 4 bits, takes one to two gigabytes of memory and fits on a Pro 3 next to your application. For a model with 7 or 8 billion parameters, choose Pro 4 or larger.
Which size should I choose?
Pro 1 for a personal agent, a bot or a few automations; Pro 2 for an agent that drives a browser, or for n8n; Pro 3 for an application with its vector database, or a compact model. Moving your server up to the next size is your call, and your data stays right where it is.
Can the AI VPS be managed?
Yes. From CA$27.99 per month with Pro 1, we take charge of the server and follow your agent proactively: we spot what is building up and bring you our recommendation before the problem shows.
Who designed the AI VPS?
Simon Vasconcelos Lee, founder of Simafri and artificial intelligence expert, who runs his own agents on our servers.
Tell us about your agent
Walk us through what your agent has to do: we choose its size with you and get your server running.