Deploying agents on a server, the simple way

An agent that runs on your laptop is a demo. An agent that runs on a server, restarts itself after a crash, and keeps working while you sleep is a system. The gap between the two is deployment, and in 2026 the path is far simpler than it looks: almost everything reduces to Docker plus one configuration file. This guide keeps it that simple.

Why Docker is the one tool that matters

The oldest problem in software is "it works on my machine". Your agent needs Python 3.12, four libraries, a model server, and a database, and the server has none of them. Docker solves this by packaging the application and everything it needs into a container: a sealed box that runs identically anywhere Docker is installed. You build the box once; the server just runs it. That is the entire idea, and it is why Docker is the default skill to learn before any other deployment tool.

The supporting cast

Docker Compose

Your whole stack in one file

What it does Describes every service (agent, model, database) in one YAML file

Why it matters One command starts, stops, and restarts everything together

Portainer

Docker without the terminal

What it does A browser dashboard over your containers: logs, restarts, updates

Why it matters Lets non-terminal people operate the server safely

Ollama

The model server in the stack

What it does Serves your local model on port 11434 for every agent to call

Why it matters Runs as a container next to the rest, GPU passed through

n8n (self-hosted)

The workflow engine, also a container

What it does Runs your visual agent workflows on the server, on schedule or on trigger

Why it matters The non-coder's whole orchestration layer in one image

A full agent stack in one file

Here is the smallest honest example: a model server and a workflow engine, wired together, surviving reboots. Save it as docker-compose.yml on the server and run one command. This is genuinely how small the core of a deployment is.

services:
  ollama:
    image: ollama/ollama
    volumes: ["ollama:/root/.ollama"]
    restart: unless-stopped

  n8n:
    image: n8nio/n8n
    ports: ["127.0.0.1:5678:5678"]
    environment:
      - N8N_BASIC_AUTH_ACTIVE=true
    volumes: ["n8n:/home/node/.n8n"]
    restart: unless-stopped

volumes:
  ollama:
  n8n:

docker compose up -d

# see what is running
docker compose ps

# read the logs of one service
docker compose logs n8n -f

Note the two quiet safety choices in that file: restart: unless-stopped brings every service back after a crash or reboot, and the n8n port is bound to 127.0.0.1, so it is reachable only from the server itself, not from the internet. You then access it through a VPN or an SSH tunnel. Those two lines prevent the two most common beginner disasters.