Skip to content
Daleel Web

AI agents & automation

Best VPS for n8n and AI agents: how much server you actually need

How much RAM and CPU n8n, OpenClaw, Hermes Agent and Ollama really need, which VPS plan size fits each, and what to check before you pick a provider.

By Daleel Web Editorial TeamPublished 2 min read
Stacked layers illustrating virtual servers on one physical machine
On this page

Most AI agent tools are light. The heavy part, the language model, usually runs in the cloud and you call it through an API. What your VPS really runs is the automation tool or agent, a database, and a reverse proxy for HTTPS. That changes the sizing question completely.

What each tool needs

ToolWhat runs on your serverPractical starting size
n8nNode.js app + SQLite or PostgreSQL2 GB RAM minimum, 4 GB comfortable
OpenClawNode.js gateway + chat channelsSmall VPS; no official RAM minimum published
Hermes AgentPython app + gatewaySmall VPS; the README says it runs on a "$5 VPS"
Ollama (local models)The language model itselfDepends on the model; far more RAM than the others

Our own measurement: n8n 2.47.7 used about 740 MB of RAM sitting idle with no workflows. Everything you add (AI agent workflows, code nodes, parallel executions, other containers) comes on top of that.

Ollama is the exception. Running the model on your own server means the model has to fit in memory. Ollama's FAQ notes that required RAM also grows with the number of parallel requests and the context length. If you want local models, plan for a much larger plan than for n8n alone, and expect slower answers on a CPU-only VPS than from a cloud API.

Three sensible sizes

  1. Starter (one tool, light use): 1–2 vCPUs, 4 GB RAM. Enough for n8n or a personal agent with a cloud model.
  2. Comfortable (n8n + agent + database): 2 vCPUs, 8 GB RAM. Room for several containers, PostgreSQL and growth.
  3. Local models: 4+ vCPUs and 16 GB RAM or more, and accept that CPU-only inference is slow.

What to check before you choose a provider

  • Server location: pick a data center close to your users and to the services your workflows call most.
  • Root access and Docker: almost every agent guide assumes Docker. A ready Docker image saves setup time.
  • Snapshots and backups: you will break something while experimenting. One-click snapshots make that painless.
  • Resizing: start small and move up a plan without reinstalling.
  • Refund window: useful while you test whether self-hosting suits you.

Where Hostinger fits

Hostinger's KVM plans line up neatly with the sizes above. According to Hostinger's VPS page:

PlanvCPURAMNVMe diskBandwidth
KVM 114 GB50 GB4 TB
KVM 228 GB100 GB8 TB
KVM 4416 GB200 GB16 TB
KVM 8832 GB400 GB32 TB

It also offers ready templates for n8n (including a queue-mode template for heavier use) and for Docker with Traefik, and states a 30-day money-back guarantee. We run this site on a KVM 2 next to other Docker projects. Check current prices and specs on Hostinger's site before buying; they change.

For a side-by-side look at Hostinger, Hetzner and DigitalOcean, read our n8n hosting comparison. Ready to set up? Follow how to install n8n on a VPS.

Frequently asked questions

Is a 1 GB or 2 GB VPS enough for n8n?

2 GB can work for light use, but in our test n8n alone used about 740 MB while idle. With Docker, a reverse proxy and a few active workflows, 4 GB gives you room to breathe.

Do I need a GPU to run AI agents?

Not if the agent uses a cloud model through an API, which is how most people start. You only need serious hardware if you want to run the language model itself on your server with a tool like Ollama.

Sources & references

  1. Hostinger: VPS hosting plans
  2. Hostinger: Self-hosted n8n
  3. Hermes Agent README (Nous Research)
  4. Ollama FAQ