---
title: "Deploy Beszel"
description: "Dashboard that tracks CPU, memory and disk across your servers"
category: "Observability"
url: https://railway.com/deploy/beszel-hub
---

# Deploy Beszel

Dashboard that tracks CPU, memory and disk across your servers

**[Deploy Beszel on Railway](https://railway.com/template/beszel-hub)**

- **Creator:** A3A
- **Category:** Observability

## Template content

### beszel https://cdn.jsdelivr.net/gh/homarr-labs/dashboard-icons/svg/beszel.svg

- **Source:** https://github.com/gridalpha/beszel-railway
- **Public domain:** Yes

## Buckets

- **beszel-backups**

## Documentation

# Deploy and Host Beszel on Railway

Beszel is an open-source server monitoring platform for people who want real machine
metrics without running Prometheus and Grafana to get them. A single hub serves the
dashboard and stores history, while a tiny agent on each monitored machine reports CPU,
memory, disk, network, load average, temperature, GPU, S.M.A.R.T. health and per-container
Docker or Podman statistics. The agent idles at roughly 10–15 MB of RAM, which is why
homelab operators and small platform teams reach for it instead of heavier observability
suites; Netdata's agent typically wants 200–500 MB for similar ground.

Self-hosting Beszel on Railway gives you the hub — the half that wants a stable public
address, a persistent database and someone else's uptime. This template runs the hub from
the `gridalpha/beszel-railway` source repository, which layers the upstream
`henrygd/beszel` binary onto Alpine and starts it on port 8090 behind a generated
`*.up.railway.app` domain. A 5 GB volume at `/beszel_data` holds the embedded PocketBase
database and the hub's signing key, and a managed object-storage bucket receives scheduled
backups of it. Your own servers stay where they are: agents dial out over an authenticated
WebSocket, so nothing on the monitored side needs an open inbound port.

![Diagram of the Beszel hub service and its volume on Railway](https://res.cloudinary.com/rroe4rtk/image/upload/v1787777081/beszel-architecture.png)

## Getting Started with Beszel on Railway

Set `USER_EMAIL` and `USER_PASSWORD` before the first deploy — those two create the first
account, and leaving them blank makes the hub open an unauthenticated first-run form to
whoever finds the URL. When the deploy is green, open the public domain and sign in.
You land on **All Systems**, an empty table waiting for its first machine.
Click **Add System**, name it, and copy either the generated `docker-compose.yml` from the
Docker tab or the one-line installer from the Binary tab; both carry the hub URL, the
registration token and the hub's public key filled in. Run it on the server you want to
watch and within a minute the row turns green and starts drawing charts.
Click the system name for CPU, memory, disk I/O, network and Docker container history over
the last hour, day, week or month. Use the bell icon on any row to set alerts — status
changes, or CPU above 80% for ten minutes — and add a Discord, Slack, Gotify or ntfy
webhook under notification settings so they reach you.

![Beszel All Systems table listing a monitored server's live metrics](https://res.cloudinary.com/rroe4rtk/image/upload/v1787777084/beszel-all-systems.png)

![Beszel charts of CPU, memory and Docker container usage](https://res.cloudinary.com/rroe4rtk/image/upload/v1787777085/beszel-system-charts.png)

![Beszel alerts panel with status and CPU threshold rules enabled](https://res.cloudinary.com/rroe4rtk/image/upload/v1787777090/beszel-alerts.png)

## About Hosting Beszel

Beszel splits cleanly in two. The hub is a single Go binary with an embedded PocketBase
database and a compiled-in React dashboard; it owns users, history, alert rules and agent
connections. The agent is a second small binary that reads the machine it runs on and
streams samples back. Because the hub keeps everything in one SQLite file on a volume, there
is no separate database, cache or worker tier — which is why it stays cheap. Self-host it
when you want per-machine history you own, across servers in different places: a VPS, a NAS
at home, a build box under a desk.

Key features:

- Historical charts for CPU, memory, disk, network, load average, temperature, fan speed,
  GPU (Nvidia, AMD, Intel), battery and S.M.A.R.T. disk health
- Per-container Docker and Podman statistics, including logs and inspect output
- Threshold and status alerts by email or webhook
- Multi-user accounts with per-system sharing, plus OAuth2 and OIDC providers
- Systemd service monitoring, and systems declared in `config.yml`

## Why Deploy Beszel on Railway

Railway removes the parts of hosting a monitoring hub that are not monitoring.

- HTTPS and a public domain provisioned automatically, certificates renewed
- A persistent volume, so history and the hub's signing key survive redeploys
- Managed object storage for off-box backups, wired in already
- One service, one replica — no orchestration to design
- Redeploys pull the current upstream release, leaving your data alone

## Common Use Cases

- **Fleet health for a small hosting setup** — watch a handful of VPS boxes from one
  dashboard and hear about a disk crossing 90% before customers do.
- **Homelab and NAS monitoring** — track temperature, fan speed and S.M.A.R.T. status on
  hardware with no vendor dashboard worth using.
- **Docker host visibility** — see which container on a crowded box is eating memory, with
  history rather than a point-in-time `docker stats`.
- **Post-incident review** — walk back through a week of load average and network throughput
  to see what changed before an outage.

## Dependencies for Beszel

- **Beszel hub** — `henrygd/beszel` (currently 0.18.8), built from
  [`gridalpha/beszel-railway`](https://github.com/gridalpha/beszel-railway). Serves the
  dashboard, REST API and agent WebSocket endpoint on port 8090.
- **Volume at `/beszel_data`** — the PocketBase database and the hub's ed25519 key. Without
  it every redeploy starts from an empty dashboard.
- **Object storage bucket** — receives scheduled backups. Leave the `S3_*` variables unset
  to skip it.

### Environment Variables Reference

| Variable | Purpose |
|---|---|
| `USER_EMAIL` | Email for the first account, created on first boot |
| `USER_PASSWORD` | Password for that account; change it in the UI later |
| `APP_URL` | Public URL used in alert links |
| `PORT` | Hub listening port, `8090` |
| `BACKUP_CRON` | Backup schedule, default `0 4 * * *` |
| `BACKUP_KEEP` | Backups retained, default `7` |
| `MFA_OTP` | Set `true` to require one-time codes at sign-in |
| `SHARE_ALL_SYSTEMS` | Set `true` to let every user see every system |

### Deployment Dependencies

- Source repository: 
- Upstream project: , documentation at 
- Images: `henrygd/beszel` (hub), `henrygd/beszel-agent` (agent)

## Hardware Requirements for Self-Hosting Beszel

| Resource | Minimum | Recommended |
|---|---|---|
| CPU | 0.25 vCPU | 0.5 vCPU |
| RAM | 128 MB | 512 MB |
| Storage | 1 GB volume | 5 GB volume |
| Runtime | Linux container, port 8090 | Same, with a persistent volume |

Storage grows slowly: the hub rolls one-minute records into ten-minute, hourly and daily
averages.

## Self-Hosting Beszel with Docker

To run the hub outside Railway, this Compose file is the whole of it:

```
services:
  beszel:
    image: henrygd/beszel:latest
    restart: unless-stopped
    ports:
      - "8090:8090"
    volumes:
      - ./beszel_data:/beszel_data
    environment:
      APP_URL: https://beszel.example.com
      USER_EMAIL: you@example.com
      USER_PASSWORD: change-this-first
```

On each monitored machine, run the agent with the token and public key the Add System
dialog gives you:

```
docker run -d --name beszel-agent --restart unless-stopped \
  --network host \
  -v /var/run/docker.sock:/var/run/docker.sock:ro \
  -e HUB_URL=https://beszel.example.com \
  -e TOKEN= \
  -e KEY="" \
  henrygd/beszel-agent:latest
```

Only one agent registers per machine: agents fingerprint the host, so a second container on
the same box is refused.

## How Much Does Beszel Cost to Self-Host?

Beszel is free and open source under the MIT licence — no paid tier, no seat count, no
metered ingestion — so the only cost is the infrastructure it runs on. On Railway that is
one small service plus a volume and a backup bucket, at the low end of usage-based billing.
Agents run on machines you already pay for.

## FAQ

**What is Beszel?**

An open-source server monitoring platform with a web hub and lightweight agents. It records
CPU, memory, disk, network, sensor and Docker container metrics from any number of machines
and alerts when a threshold is crossed.

**What does this Railway template deploy?**

The Beszel hub — dashboard, REST API and agent endpoint — on a public HTTPS domain, with a
volume for its database and a managed bucket for scheduled backups. You install agents
yourself on the machines you want to monitor.

**Why does the template include a volume and an object-storage bucket?**

Beszel stores users, alert rules, history and the key that authenticates agents in one
embedded database file. The volume keeps that file across redeploys; the bucket holds
nightly copies off the box.

**How do I add a server to a self-hosted Beszel hub?**

Sign in, click **Add System**, and copy the generated Docker Compose file or binary
installer. Both include the hub URL, a registration token and the hub's public key. The
agent connects outward, so no inbound firewall rule is needed. For a fleet, enable a
universal token and agents self-register.

**Can I run the Beszel agent on Railway too?**

Not usefully. The agent measures the whole machine it runs on, and inside a container that
reads the shared host rather than the container's own limits, so the numbers describe
nothing you own. Agents belong on the servers you are monitoring.

**How do I send Beszel alerts to Slack or Discord instead of email?**

Beszel uses Shoutrrr URLs. Under notification settings add a webhook for Discord, Slack,
ntfy, Gotify or a generic HTTP endpoint, then enable the alerts you want per system. Email
works once SMTP is configured.


## Similar templates

- [Pyroscope profiling](https://railway.com/deploy/pyroscope-profiling) — Protected continuous profiling with durable Pyroscope storage.
- [SigOnly](https://railway.com/deploy/sigonly) — Deploy SigNoz with a working demo app & config in one click
- [Unwrapped Spotify Music Stats](https://railway.com/deploy/wide-bold) — Unwrapped Spotify Music Stats, Estatísticas de músicas disponíveis

Open this page in a browser: https://railway.com/deploy/beszel-hub