AI agents
Long-running agents and assistants that call a hosted model through your own API key.
Use case · AI
AI VPS: run AI agents and tools on a CPU server
Yes, you can run AI on a VPS with no GPU: the model does its heavy work on OpenAI's or Anthropic's servers through your own API key, while the VPS runs the agent, the workflow or the tool that calls it. RyzenRDP is a CPU VPS for API-driven AI and self-hosted AI tools — no GPU, and no model training.
- CPU-only: no GPU, no model training
- Your own OpenAI or Anthropic API key
- Linux (root over SSH) or Windows Server
What you can run on an AI VPS
n8n automation
Self-hosted n8n workflows that call AI APIs, move data between apps and run on a schedule.
MCP servers
Model Context Protocol servers that give an assistant tools and data over a long-running connection.
AI-powered bots
Discord and Telegram bots that call AI APIs to answer, summarise or moderate messages.
Retrieval-augmented generation (RAG) with a vector database such as Qdrant or pgvector runs on the same servers; the AI agent hosting page covers that setup. Everything here is CPU work — the heavy model math runs through an API, or in a small local model that runs slowly on the CPU.
Do you need a GPU to run AI on a VPS?
No — most AI apps do not need a GPU on the server. When your agent, bot or workflow calls a hosted model through an API (OpenAI, Anthropic and others), the GPU work happens on that provider's servers. Your VPS only runs the code that builds the prompt, calls the API and handles the reply, which is ordinary CPU and memory work that AMD Ryzen cores handle well.
A VPS has no GPU, so it cannot train models, generate images on the server or run large local models at useful speed. If that is what you need, a CPU VPS is the wrong tool.
Can I run a local LLM on a CPU VPS?
You can run small, quantized models on the CPU, but they are slow. A 7B-parameter model in GGUF format, served with llama.cpp or Ollama, will load and answer on a CPU-only server — expect a few tokens per second, not the fast responses you get from a hosted API. It suits background jobs, drafts and testing rather than a busy chat endpoint.
Size the plan for the model: a 7B model quantized to 4-bit needs roughly 5–8 GB of RAM, so start at T3 (6 vCPU, 8 GB RAM and 160 GB Gen4 NVMe SSD) or larger. For anything bigger, or for real-time replies, call a hosted model through its API instead.
Which plan for which AI workload
Shown in Dallas. T1 runs one API-driven agent, bot or MCP server. T2 fits an agent plus n8n and a few tools, or several API-driven services at once. T3 adds room for a vector database and retrieval, or a small quantized local model on the CPU. Need more memory for a larger local model? Larger tiers go up to 32 GB of RAM — see all plans.
T1 · Dallas
AMD Ryzen
$14.99/mo
USD per month · billed for 1, 3, 6 or 12 months
- 2 shared vCPU
- 2 GB DDR5 RAM
- 40 GB Gen4 NVMe SSD
- 1 Gbps port · 1 IPv4
Room for an agent and n8n
T2 · Dallas
AMD Ryzen
$24.99/mo
USD per month · billed for 1, 3, 6 or 12 months
- 4 shared vCPU
- 4 GB DDR5 RAM
- 80 GB Gen4 NVMe SSD
- 1 Gbps port · 1 IPv4
T3 · Dallas
AMD Ryzen
$39.99/mo
USD per month · billed for 1, 3, 6 or 12 months
- 6 shared vCPU
- 8 GB DDR5 RAM
- 160 GB Gen4 NVMe SSD
- 1 Gbps port · 1 IPv4
Plan prices do not go up at renewal, and there is no setup fee. Prices and specs from our live catalog, checked .
Linux-first, with Windows Server available
AI tools are built for Linux first, so a RyzenRDP VPS runs Ubuntu, Debian or AlmaLinux, reached as root over SSH.
You install what you need yourself: apt for packages, Docker for n8n or a vector database, npm
for Node tools, and a systemd service so an agent restarts after a reboot. You choose the versions
and keep your data and logs on your own disk.
Windows Server is available too, at order or on reinstall, if your tooling needs it. Either way you get full root or Administrator access, a 1 Gbps port and one dedicated IPv4 address.
What is not allowed
No spam or bulk messaging, no scraping that breaks a site's terms, no fake accounts or engagement, and no using a model to generate abusive, deceptive or illegal content. Automation must follow the rules of the services it touches. See the acceptable use policy.
AI VPS FAQ
Do I need a GPU to run an AI agent?
No. When your agent calls a hosted model through an API (OpenAI, Anthropic and others), the GPU work runs on the provider's servers. Your VPS runs the agent code, which is CPU and memory work. No plan includes a GPU.
Can I run a local LLM on this VPS?
You can run small, quantized models on the CPU, but slowly. A 7B model in GGUF format via llama.cpp or Ollama needs about 5–8 GB of RAM, so start at T3 (6 vCPU, 8 GB RAM and 160 GB Gen4 NVMe SSD), and returns a few tokens per second. For larger models or fast replies, call a hosted model through its API.
Is an AI VPS for GPU AI like image generation or training?
No. There is no GPU, so the server cannot train models, generate images or run large local models at useful speed. It is built for API-driven AI and small CPU-run tools.
How much RAM does n8n need?
A self-hosted n8n instance with a handful of workflows is comfortable on T2 (4 vCPU, 4 GB RAM and 80 GB Gen4 NVMe SSD). Many workflows, or ones that hold large payloads in memory, want T3 (6 vCPU, 8 GB RAM and 160 GB Gen4 NVMe SSD) or more.
Can I host an MCP server on a VPS?
Yes. A Model Context Protocol server is an ordinary long-running process: install it with npm or Docker, run it as a systemd service, and keep its port behind a firewall rule or reachable only over SSH. See MCP server hosting.
Do you install the AI tools for me?
No. You get root over SSH, or Administrator on Windows, and install what you need yourself — apt, Docker, npm and systemd. You choose the versions and keep your data and logs on your own disk.
Which operating system should I use for AI work?
Linux. Most AI tools target Ubuntu, Debian or AlmaLinux first, reached as root over SSH, and Docker images assume Linux. Windows Server is available at order or on reinstall if your tooling needs it.
Can I build a RAG app with a vector database?
Yes. Run a vector database such as Qdrant, Chroma or Postgres with pgvector in Docker, and your agent queries it for retrieval. Sizing and setup are on the AI agent hosting page.
Deploy your AI VPS
A CPU VPS for API-driven AI and self-hosted tools, on AMD Ryzen. Plans from $14.99/mo.