Self-Host n8n and OmniRoute on a $5 VPS: The Complete 2026 Guide

Why Your Automation Dies When Your Laptop Sleeps

You built a workflow in n8n, it runs beautifully — right up until you close your laptop. Then the agent stops, the webhook goes silent, and the automation you spent an evening wiring up is dead until you wake the machine back up. This is the exact problem a VPS solves.

A VPS (virtual private server) is simply a computer that stays online 24/7. Deploy your automation there and it keeps running whether your laptop is on, off, or in another country. In this guide I walk through the full stack from my video: n8n for the workflows, OmniRoute for free AI model access, and AgentRouter for $175 in Claude credit — all on a ~$5/month server.

What You're Building

Three pieces, one server:

n8n — the AI agent workflow builder. You drag nodes on a canvas, assign agents, and deploy automations that reason and act.
OmniRoute — an open-source gateway to 339+ AI providers (90+ of them free) behind one OpenAI-compatible endpoint. Its killer feature is automatic fallback: when one model's tokens run out, it switches to a free or alternate provider without breaking your workflow.
AgentRouter — one API key that unlocks Claude Opus, GPT, DeepSeek and more, with $175 in free credit when you sign up through my link.

By the end, n8n talks to OmniRoute, OmniRoute routes to free and paid models, and everything runs 24/7 on hardware you control.

Step 1 — Get a Contabo VPS With One-Click n8n

I use Contabo because they offer a one-click n8n application install for as low as $5.28/month. Pick the Cloud VPS 6 (or VPS 12) — the smaller VPS 4 runs the stack fine for testing, but as you deploy more automations it becomes an overload, and upgrading later is more technical than starting one tier up.

Choose your region: EU or US. Contabo recommends United Kingdom for best latency; EU is free to select.
Generate a root password during setup — copy it somewhere safe.
Fill in your details, pay with PayPal or card, and pick the 1-month plan to start.

Within minutes you get an email with your login data: the server IP, the n8n application URL, the auto-generated username, and the password you set.

Step 2 — First Login and License Activation

Open the n8n address from the email, sign in with the generated credentials, and claim your free license key. Use your own email (not the default one) when requesting the key, then paste the activation code under Usage and Plan → Enter activation key. Your n8n instance is now live and activated on your server.

Step 3 — SSH In and Install Docker

Everything else runs through Docker, so open a terminal and connect to your server:

ssh root@YOUR_VPS_IP

Then install Docker with the official convenience script:

curl -fsSL https://get.docker.com | sh

Rule of thumb for the whole guide: if you can install it with Docker, do it with Docker. It keeps services isolated and restarts them automatically after reboots.

Step 4 — Install OmniRoute

Head to the official OmniRoute docs and use the Docker guide's quick run — it sets sane default credentials for you. The equivalent Docker Compose setup looks like this:

services:
  omniroute:
    image: diegosouzapw/omniroute:latest
    container_name: omniroute
    restart: unless-stopped
    ports: ["20128:20128"]
    environment: {PORT: 20128, LOG_LEVEL: info}

docker compose up -d

When the logs show it listening on port 20128, OmniRoute is installed on your server.

Step 5 — Onboard the OmniRoute Dashboard

Open http://YOUR_VPS_IP:20128 in your browser. The onboarding wizard asks you to set a dashboard password, optionally pick a provider (skip it for now), and run a connection test (skip). The important output is your endpoint URL — copy it, you'll paste it into n8n in Step 8.

Step 6 — Claim $175 in Claude Credit via AgentRouter

This is the free-tier unlock. Go to AgentRouter and sign up — use "Continue with GitHub," not email. One catch: your GitHub account needs to be at least 3 months old to qualify for the full credit.

Once inside, create an API key (name it anything, toggle it to unlimited) and copy it. Then in OmniRoute: Providers → search "AgentRouter" → Add connection → paste the key → validate → save. OmniRoute imports the available models — Claude Opus, Sonnet, and the rest. Test one in the built-in chat bot to confirm it responds.

Step 7 — Add Free Providers (26 of Them)

OmniRoute has a Free Tier section listing roughly 26 platforms whose APIs you can connect at zero cost. Connect as many as you like — this is your fallback pool when paid credits run dry, and it's what makes "no API bill" realistic rather than a slogan.

Step 8 — Connect n8n to OmniRoute

Two small moves wire the whole stack together:

In OmniRoute, go to API Keys → Create API key (name it "n8n") and copy it.
In n8n, create a new credential of type OpenAI, paste the OmniRoute API key, and replace the base URL with your OmniRoute endpoint (the one you copied in Step 5). Save.

n8n now believes it's talking to OpenAI — but every request flows through OmniRoute's routing and fallback layer.

Step 9 — Build Your First AI Agent Workflow

Create a workflow with a Chat Trigger and an AI Agent node. In the agent's chat model, select your OmniRoute credential and pick a model (Claude Opus works well). Open the chat, send a message, and watch it respond — your agent is running on a free-routed model, on your own server, 24/7.

Step 10 — The Combo Feature (Stop Switching Models Manually)

Here's the part most people miss. Without combos, when one model's tokens run out you have to manually switch your n8n node to another model — and if you're not around, your automation breaks. OmniRoute's Combo fixes this: create a combo (e.g. name it "free"), select multiple models — AgentRouter's Claude Opus, DeepSeek, whatever you've connected — and OmniRoute exposes the combo as a single model choice. Point n8n at the combo and it silently fails over between members. No manual switching, no broken automations.

What It All Costs

Contabo Cloud VPS 6: ~$5–7/month
OmniRoute: free, open source
AgentRouter: $175 free credit via GitHub signup
Free-tier providers: $0

Total: about the price of a lunch, for an always-on AI automation stack.

Next Steps

The full command list from the video is in the description, and these guides go deeper:

And if you want the whole thing explained visually, the 25-minute walkthrough is embedded above.