HELPER VPS

AI agent VPS: AI automation servers

The HELPER VPS AI agent VPS is a Linux cloud server for running agents that call LLM APIs, n8n workflows and chatbots 24 hours a day in 40 cities, starting at $25.00 per month. On memory optimized plans with 16 GB to 256 GB of memory, you can also run small open models yourself with Ollama.

Which plan should I choose for an LLM VPS?

The specs AI work needs depend on where the model runs. If you call external LLM APIs, a small server is enough; if the model runs on the server, you need as much memory as the model size.

    To run models on the server, allow generous memory beyond the model file size, and for large models that need fast responses, consider a GPU server as well.

    What can I run on an AI automation server?

    A VPS is a computer that never switches off, so agents keep working without your laptop staying on. Here are five common setups.

    Why run AI agents on a VPS?

    Tools used alongside AI servers

    Most AI tools officially support Linux and Docker. Create the server with the Docker app image and you can run n8n, Ollama and vector databases as containers right away.

    Place your server close to your users

    Choose the location closest to your users among 40 cities in 31 countries.

    How fast is running an LLM on a CPU-only VPS?

    Running models on CPU produces answers more slowly than on GPU. 3B to 8B quantized models are readable at chat-like speed, but generating long text at once or serving many people simultaneously means longer waits. So local models on CPU servers fit best for work producing lots of short answers, such as classification, short summaries and extracting information.

    Ollama's guidance recommends 8 GB of memory for 7B models, 16 GB for 13B models and 32 GB or more for 33B models. The operating system and other programs use memory too, so choose one plan up, and it is best to check quality and speed with a small model first and then scale up.

    Services needing fast responses, large models of 70B or more, image generation and fine-tuning need a GPU. Look into these on the GPU servers page, and split roles so the VPS handles agents and workflows that call GPU servers or LLM APIs, balancing cost and speed.

    How do I install n8n and Ollama on a VPS?

    Choose the Docker app as the image when creating the server and it starts with Docker Engine and Docker Compose installed on Ubuntu 24.04. After connecting, create storage with docker volume create n8n_data and launch n8n with docker run -d --name n8n -p 5678:5678 -v n8n_data:/home/node/.n8n docker.n8n.io/n8nio/n8n. For production, connecting a domain and HTTPS is safer.

    Install Ollama with one line, curl -fsSL https://ollama.com/install.sh | sh, then enter a model name like ollama run llama3.2 to download and run it right away. Once installed, an API opens on port 11434 within the server, and n8n or Python code on the same server calls this address to use the local model.

    By default, the Ollama API accepts connections only from within the server, and it is best to leave it that way. If you need to use it from outside, come in through a VPN such as WireGuard, or put authentication on an Nginx proxy so only allowed people can use it.

    How do I run an AI agent server securely?

    Do not write LLM API keys directly in code; keep them in environment variables or files with narrow permissions. Set usage limits on the API side so costs do not suddenly jump even if an agent gets stuck in a loop. Connect to the server only with SSH keys, and protect admin interfaces such as n8n with strong passwords and HTTPS.

    When agents use tools that read the web and run commands, giving them narrow permissions matters. Create a separate Linux user for the agent and allow only the folders it needs, or run it inside a Docker container, so mistakes do not spread to the whole server.

    Send the agent's alert emails through a mail delivery service on port 587, and Telegram or Discord webhooks work well alongside. With the automatic backup option on, workflows and chat history are kept as well, and when you need a restore, request it on the support page after your first payment.

    More VPS by use case

    AI agent VPS: AI automation servers FAQ

    Do I really need a GPU to run AI agents?

    For agents that call LLM APIs, a small VPS without a GPU is enough. Model computation happens on the API side, and the server only handles requests and results. Consider a GPU server only when you want to run large models yourself quickly.

    How large a model can I run with Ollama?

    On memory optimized plans, you can run models that fit your memory size. 16 GB handles 7B to 8B quantized models, 32 GB around 13B, and 64 GB or more around 30B, but since they run on CPU, answers come more slowly than on a GPU.

    Is an LLM running on CPU usable?

    For work producing short answers, it is perfectly usable. Tasks with short results, such as document classification, summaries, extracting information and creating embeddings, run well on CPU, while services generating lots of long text in real time are better suited to GPU servers.

    Which plan should I run n8n on?

    n8n for individuals or small teams runs comfortably on 4 to 8 GB of memory. If you have many workflows or handle large files, choose an NVMe plan with 12 GB or more and deploy it with Docker following the install guide on the n8n app page.

    Can I call the OpenAI API or other LLM APIs from a VPS?

    Yes, outbound HTTPS requests from the server are free to use. Keep API keys in the server's environment variables and call them from your agent or n8n; API usage rules follow each service's terms.

    How many AI automation servers can I create?

    You can create up to 10 servers per account. You can split agent servers, database servers and local model servers, or keep a separate chatbot server for each customer region.

    Can agents send email?

    Yes, by connecting a mail sending service that uses port 587. Every server keeps port 25 closed to protect IP reputation, so put the mail sending service's SMTP details or API into your agent.

    What if model files are large and the disk runs short?

    Attach an extra disk of 50 to 1000 GB when creating the server for plenty of room for model files and data. Memory optimized plans use NVMe disks, so models load quickly too.

    How do I pay for AI servers?

    Your server is created for the billing period you choose (1, 3, 6 or 12 months). The minimum payment is $25.