Your First Homelab with PodWarden

You have a spare PC, or maybe an old workstation collecting dust. You've heard people talk about homelabs — running their own cloud storage, media servers, ad blockers, even local AI models. It sounds useful, but the learning curve looks steep: Proxmox, Docker, Kubernetes, YAML files, networking, reverse proxies.

Here's the short version: you don't need most of that to start. One machine, Ubuntu, and PodWarden gets you from bare hardware to running services in under an hour. No Kubernetes knowledge required. No cloud subscriptions. No ongoing costs.

PodWarden is free for this. Not free-with-limits, not free-for-30-days. The self-hosted tool is free for unlimited servers, unlimited deployments, unlimited workloads. It runs entirely on your hardware.

Why People Build Homelabs

People self-host for two reasons: control and cost.

You become the custodian of your own data. Your photos aren't in Google's cloud. Your passwords aren't in someone else's database. Your documents aren't on a server you don't control. When a cloud provider changes their terms, raises prices, or shuts down a product, it doesn't affect you. Your data is on your hardware, in your home, under your control.

You stop paying rent on software you could own. Cloud storage, password managers, media streaming, note-taking apps, smart home platforms — these add up. A homelab replaces dozens of monthly subscriptions with software you run yourself, on hardware you already have. The electricity cost for a single machine is a few dollars a month. The software is open-source and free.

There's a third reason people don't always admit: it's fun. Homelabs are a playground for learning real infrastructure skills — networking, Linux, containers, automation — in a space where mistakes don't cost you your job.

What You Need

Hardware

Any PC from the last decade works. A typical starter machine looks like this:

Old office PCs, decommissioned desktops, mini PCs, used workstations — all work. If it can run Ubuntu, it can run a homelab. Laptops with broken screens make surprisingly good headless servers.

If you have an NVIDIA GPU — even an older one with 6–8 GB VRAM — that's a significant bonus. PodWarden detects it automatically and you can run local AI models that would cost real money on cloud APIs.

Software

You need exactly two things installed:

1. Ubuntu Server 22.04 or 24.04 — free, well-documented, and the most widely supported Linux distribution
2. Docker and Docker Compose — the container runtime (install Docker)

That's the entire foundation.

Do I Need Proxmox? Virtual Machines? Kubernetes?

No, no, and no — not to start.

This is where most homelab guides lose people. They tell you to install a hypervisor, create VMs, learn Kubernetes, set up a cluster — before you've deployed a single useful application. That's backwards.

Proxmox is a hypervisor for running multiple virtual machines on one physical machine. It's useful when you need to run completely separate operating systems (like a Windows VM alongside Linux, or TrueNAS as a dedicated NAS OS). But for a first homelab where you're running containerized applications, Proxmox adds complexity and overhead without solving a real problem. Every GB of RAM and every CPU core the hypervisor uses is a resource your actual applications can't use.

Kubernetes is a container orchestrator designed for large-scale production systems. PodWarden uses K3s (a lightweight Kubernetes distribution) under the hood, but you never interact with it directly. You don't write manifests, you don't use kubectl, you don't learn pod specs. PodWarden handles all of that. You work with servers, templates, and a deploy button.

Virtual machines make sense when you need them — and for most homelab starters, you don't. The one common exception is TrueNAS, which is a full operating system designed to manage disk arrays and needs bare-metal or VM access to your drives. Everything else on the typical homelab wishlist runs in Docker containers, managed by PodWarden.

Start bare metal. Add complexity only when you hit a wall that requires it. Most people never do.

Installing PodWarden

Install Ubuntu Server on your machine, install Docker, then run:

curl -fsSL https://www.podwarden.com/install.sh | bash

The installer is interactive. It checks prerequisites, asks a few questions (ports, auth method), generates configuration, and starts PodWarden. The whole process takes about five minutes.

Finding Your Server's IP Address

To open the PodWarden dashboard from another computer on your network, you need your server's local IP address. On the server, run:

hostname -I

This prints your server's IP addresses. The first one is typically your local network address — something like 192.168.1.50 or 10.0.0.12. That's the one you want.

If you see multiple addresses and aren't sure which one to use, ip addr gives more detail:

ip addr show

Look for the address under your main network interface (usually eth0, eno1, or enp0s3). It's the inet line that starts with 192.168., 10., or 172.16–31. — that's your LAN address.

Once you have it, open a browser on any computer on the same network and go to:

http://192.168.1.50:3000

Replace 192.168.1.50 with your actual IP. That's your PodWarden dashboard.

Tip: If you want a stable address that doesn't change, log into your router and assign a static IP (sometimes called a "DHCP reservation") to your server's MAC address. This way, your server always gets the same IP, and your bookmarks keep working. Most routers have this option under DHCP or LAN settings.

PodWarden itself is lightweight — a PostgreSQL database, a FastAPI backend, and a Next.js frontend. It uses roughly 2 CPU cores and 2 GB RAM, leaving the rest of your machine's resources available for actual workloads.

What Happens Next

From the dashboard:

1. Your machine appears as a host. PodWarden detects its hardware — CPU cores, RAM, disk, GPU model and VRAM (if you have one).
2. You provision it as a cluster. Click "Provision as Control Plane" and PodWarden installs K3s, plus NVIDIA GPU drivers if applicable. Takes a few minutes.
3. You deploy applications. Browse the PodWarden Hub catalog, import a template, click Deploy. Or create your own stacks from any Docker image.

No YAML. No terminal commands after the initial install. No Kubernetes concepts to learn.

For the full setup walkthrough — authentication, SSH keys, adding additional hosts, Tailscale integration — see Getting Started.

Built-in AI Assistant via MCP

PodWarden has a built-in Model Context Protocol (MCP) integration. This means you can connect any AI assistant — Claude, ChatGPT, a local model running on your own hardware — and use natural language to manage your entire infrastructure.

Instead of clicking through dashboards or memorizing CLI commands, you can say things like:

• "What's running on my cluster right now?"
• "Deploy Jellyfin to my home server"
• "Why is my Nextcloud pod failing to start?"
• "How much GPU memory is available?"
• "Show me the logs for Ollama"

The AI assistant has access to 41 MCP tools covering everything: host management, cluster operations, workload deployment, storage, secrets, troubleshooting, and monitoring. It can inspect your infrastructure, diagnose problems, deploy applications, and explain what's happening — all through conversation.

This isn't a gimmick. For homelab users who don't want to become Kubernetes experts, it means you can troubleshoot and configure your system by describing what you want in plain English. The AI handles the translation to infrastructure operations.

See MCP Integration for setup instructions.

What People Run on Homelabs

The PodWarden Hub catalog has over 2,000 pre-configured templates across 13 categories. Here's what homelab users typically deploy first, and why.

Ad Blocking & DNS

AdGuard Home or Pi-hole — network-wide ad and tracker blocking. Point your router's DNS to your homelab, and every device on your network gets ad-free browsing without installing browser extensions. This is the single most immediately useful homelab service. You'll notice the difference on your first page load.

Media Streaming

Jellyfin (free, open-source) or Plex (free tier available) — your own Netflix. Point it at a folder of movies, TV shows, or music, and stream to any device in your house. Jellyfin has no paywalls; Plex has wider device support and easier remote sharing. Many people run both.

Jellyseerr — a request manager that lets family members request movies and shows. Tautulli — monitoring and analytics for Plex usage.

If you have an NVIDIA GPU, Jellyfin and Plex use it for hardware-accelerated transcoding — converting video formats on the fly so any device can play any file smoothly.

Photo Management

Immich — a self-hosted Google Photos replacement. Automatic backup from your phone, facial recognition, location-based organization, shared albums. It's the most compelling reason for non-technical family members to care about your homelab. Your photos stay on your hardware, not in Google's cloud.

Cloud Storage & Documents

Nextcloud — your own Dropbox and Google Drive. File sync across devices, document collaboration, calendar, contacts, and hundreds of plugins. It's the Swiss Army knife of self-hosted software. Runs well with 1–2 GB of RAM.

Password Management

Vaultwarden — a lightweight, self-hosted Bitwarden-compatible password manager. Works with all official Bitwarden browser extensions and mobile apps. Your passwords are encrypted and stored on your machine, not in someone else's cloud. Uses almost no resources — under 100 MB of RAM.

Smart Home

Home Assistant — the hub for home automation. Controls thousands of smart home devices from a single interface. Automate lights, thermostats, cameras, locks, and sensors. The homelab connection matters here: Home Assistant runs locally, which means your smart home doesn't depend on internet connectivity or a vendor's cloud staying operational.

Local AI

Ollama — run large language models locally. With an 8 GB NVIDIA GPU, you can comfortably run 7B–13B parameter models (Llama, Mistral, DeepSeek, Phi, Gemma). Private, free, no API costs, no rate limits. With a 6 GB card, smaller models still work well.

This is where a homelab GPU really earns its keep. Cloud API calls for LLMs cost money per token. A local model running on your own GPU costs nothing after the initial hardware.

Monitoring & Dashboards

Grafana — visualization and dashboards. Prometheus — metrics collection. Loki + Promtail — log aggregation. Together, these give you complete visibility into what's running, how it's performing, and what's gone wrong. Grafana dashboards are genuinely satisfying to set up and look at.

Networking & Reverse Proxy

Traefik or Nginx Proxy Manager — reverse proxies that let you access your services via clean URLs (like jellyfin.home.local instead of 192.168.1.50:8096) and handle HTTPS certificates automatically.

Development

Gitea — a lightweight, self-hosted Git service. Your own private GitHub. Repository hosting, issue tracking, CI/CD with Gitea Actions. If you write code, this is where your projects live — on your hardware, with no repository limits.

Games

PodWarden Hub has over 150 game server templates. Minecraft, Valheim, Terraria, Satisfactory, and many more. Deploy a game server for your friends in a few clicks instead of paying for hosted game servers.

What Doesn't Run in Docker (and When You'd Need VMs)

Almost everything on the list above runs in containers. The exceptions are narrow and specific:

If you find yourself needing one of these, that's when Proxmox becomes worthwhile. Install Proxmox on the bare metal, run PodWarden in a VM with most of your resources and GPU passthrough, and run TrueNAS or Windows in a separate VM.

But don't start there. Start with bare metal Ubuntu and PodWarden. If you later need VMs, you can reinstall with Proxmox and migrate. In practice, most homelab users find that containers handle everything they need.

What This Actually Costs

Compare that to what you'd pay for cloud equivalents: Google One storage ($3–10/month), a password manager ($3–5/month), cloud photo storage ($2–3/month), a VPN ($5–10/month), game server hosting ($10–30/month), cloud AI APIs ($20+/month for regular use). A homelab pays for itself quickly.

Getting Started Today

Here's the concrete path from zero to a working homelab:

1. Install Ubuntu Server on your spare machine (Ubuntu install guide)
2. Install Docker (Docker install guide)
3. Install PodWarden — curl -fsSL https://www.podwarden.com/install.sh | bash
4. Provision your machine — open the dashboard, provision as a cluster
5. Deploy your first app — start with AdGuard Home (instant value) or Jellyfin (instant fun)
6. Connect an AI assistant — set up MCP and manage everything through conversation

You'll have a working homelab before dinner. Everything after that is exploring the catalog and adding services as you find uses for them.

Welcome to self-hosting. Your data, your hardware, your rules.