https://github.com/pouzor/homelable
Self-hosted homelab infrastructure visualizer — interactive network diagram with live status monitoring
https://github.com/pouzor/homelable
diagram network proxmox-ve self-hosted
Last synced: about 1 month ago
JSON representation
Self-hosted homelab infrastructure visualizer — interactive network diagram with live status monitoring
- Host: GitHub
- URL: https://github.com/pouzor/homelable
- Owner: Pouzor
- Created: 2026-03-06T21:49:54.000Z (4 months ago)
- Default Branch: main
- Last Pushed: 2026-04-18T21:10:28.000Z (2 months ago)
- Last Synced: 2026-04-18T21:26:37.965Z (2 months ago)
- Topics: diagram, network, proxmox-ve, self-hosted
- Language: TypeScript
- Homepage:
- Size: 2.94 MB
- Stars: 1,191
- Watchers: 4
- Forks: 41
- Open Issues: 20
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Homelable
Homelable is a self-hosted infrastructure visualization solution. It provides a network scanning feature to accelerate the identification of machines and services deployed on your local infrastructure.
Homelable also offers a healthcheck system (WIP) through multiple methods (ping/TCP, /health API, etc.) to get a global overview of online/offline services.
You can also select some pre-built design styles, or personalize each device in your diagram.
If you just like the design, you can only run the frontend and export your design as PNG.
---
## Screenshots
---
## Installation
Docker, Proxmox LXC, build from source, configuration, and development setup are all covered in **[INSTALLATION.md](./INSTALLATION.md)**.
---
## Network Scanner
The scanner runs `nmap -sV --open` on your configured CIDR ranges and populates a **Pending Devices** queue. From the sidebar you can then approve (adds a node to the canvas), hide, or ignore each discovered device.
### Triggering a scan
To save you time when mapping your infrastructure, Homlable can scan your network and report all the services it detects. It can also identify them, saving you even more time.
Click **Scan Network** in the sidebar. The Scan History tab opens automatically and refreshes every 3 seconds until the scan completes.
### macOS / root privileges
Some nmap scan types (SYN scan, OS detection) require root. If the scan fails with a permissions error, run it manually with sudo using the included script:
```bash
cd backend
sudo python ../scripts/run_scan.py 192.168.1.0/24
# Multiple ranges:
sudo python ../scripts/run_scan.py 192.168.1.0/24 10.0.0.0/24
```
Results are written directly to the database and appear as Pending Devices in the UI without restarting the backend.
> On Linux the backend process itself can be given the `NET_RAW` capability instead of running as root:
> ```bash
> sudo setcap cap_net_raw+ep $(which nmap)
> ```
---
## Node Check Methods
Homelable continuously monitors your nodes and displays their live status (online / offline / unknown) directly on the canvas. Each node can be configured with an independent check method suited to the service it runs.
| Method | Description |
|--------|-------------|
| `ping` | ICMP ping |
| `http` | GET request, success if status < 500 |
| `https` | GET with TLS verify |
| `tcp` | TCP connect (target: `host:port`) |
| `ssh` | TCP connect to port 22 |
| `prometheus` | GET `/metrics` |
| `health` | GET `/health` |
---
## Zigbee2MQTT Import
Homelable can connect directly to your MQTT broker and import your Zigbee network topology from **Zigbee2MQTT**, placing each device on the canvas as a typed node.
### Prerequisites
- A running **MQTT broker** (e.g. Mosquitto) accessible from the Homelable host
- **Zigbee2MQTT** connected to the broker with at least one device paired
### Usage
1. Click **Zigbee Import** in the left sidebar (below "Scan Network")
2. Enter your broker host, port (default `1883`), optional credentials, and base topic (default `zigbee2mqtt`)
3. Click **Test Connection** to verify reachability, then **Fetch Devices**
4. Select the devices you want from the grouped list (Coordinator / Router / End Device)
5. Click **Add N to Canvas** — devices are placed in a grid with IoT edges
### Node Types
| Type | Z2M Device | Icon |
|------|-----------|------|
| `zigbee_coordinator` | Coordinator | Network hub |
| `zigbee_router` | Router (mains-powered) | Radio |
| `zigbee_enddevice` | End Device (battery) | Antenna |
Hierarchy is set automatically: coordinator → routers → end devices (`parent_id`).
LQI (Link Quality Indicator) is stored as a node property.
> **Full documentation:** [docs/zigbee-import.md](./docs/zigbee-import.md)
---
## Live View (read-only public canvas)
Live View lets you share a read-only snapshot of your canvas with anyone on your network — no login required. It is disabled by default.
### Activation
Add LIVEVIEW_KEY to your .env:
`LIVEVIEW_KEY=your-secret-key`
Then restart the backend:
`docker compose restart backend`
### Usage
Use this URL to view your canvas:
http:///view?key=your-secret-key
The page shows your canvas in pan/zoom-only mode — no editing, no credentials needed. Clicking a node that has an IP opens it in a new tab.
---
## MCP Server (AI Integration) (optional)
Homelable can exposes a [Model Context Protocol](https://modelcontextprotocol.io) server so any MCP-compatible AI client (Claude Code, Claude Desktop, Open WebUI…) can read your homelab topology and act on it.
### What the AI can do
| | Action |
|---|---|
| **Read** | List all nodes, edges, full canvas, pending devices, scan history |
| **Write** | Add / update / delete nodes and edges, trigger a network scan, approve or hide discovered devices |
### Setup
**1. Add the keys to your `.env`:**
```env
# Authenticates AI clients (Claude Code, etc.) → MCP server
MCP_API_KEY=mcp_sk_changeme
# Authenticates MCP server → backend (internal Docker network only, never exposed)
MCP_SERVICE_KEY=svc_changeme
# Generate both with:
# python3 -c "import secrets; print(secrets.token_hex(32))"
```
No plain-text passwords involved — `AUTH_PASSWORD_HASH` is only used for the web UI login.
**2. Start the MCP service:**
```bash
docker compose up -d mcp
# MCP server is now listening on http://:8001
```
**3. Configure your AI client:**
**Claude Code** — run this command in your terminal:
```bash
claude mcp add --transport sse homelable http://:8001/mcp \
--header "X-API-Key: mcp_sk_yourkey"
```
Or add it manually to `~/.claude.json`:
```json
{
"mcpServers": {
"homelable": {
"type": "sse",
"url": "http://:8001/mcp",
"headers": {
"X-API-Key": "mcp_sk_yourkey"
}
}
}
}
```
**Claude Desktop** — edit `~/Library/Application Support/Claude/claude_desktop_config.json` (macOS) or `%APPDATA%\Claude\claude_desktop_config.json` (Windows):
```json
{
"mcpServers": {
"homelable": {
"type": "sse",
"url": "http://:8001/mcp",
"headers": {
"X-API-Key": "mcp_sk_yourkey"
}
}
}
}
```
### Example prompts
- *"What nodes are currently offline?"*
- *"Add a new LXC container named `pihole` at 192.168.1.5, connected to my switch."*
- *"Trigger a network scan on 192.168.1.0/24 and show me the pending devices."*
- *"Show me the full canvas topology."*
### Security
- The MCP server is **not** intended to be exposed to the internet — keep port 8001 firewalled to your LAN.
- Rotate the key any time by updating `MCP_API_KEY` in `.env` and restarting: `docker compose restart mcp`.
- The MCP server communicates with the backend over the internal Docker network — the backend API is never directly exposed to MCP clients.
---