Monitoring without building a monitoring department
Beszel occupies the useful middle between a terminal system monitor and a full observability platform. It collects host and container measurements through a small agent, stores history in a central hub, and presents the result in a clean web interface. The hub is built on PocketBase, while agents run on the machines being watched. That split gives you one place to inspect a fleet without putting the dashboard and database on every server.
The metric coverage is unusually practical for homelabs and small infrastructure. Beszel tracks CPU, memory, swap, ZFS ARC, disks, disk I/O, network traffic, load, temperature, batteries, and Docker or Podman containers. It can also collect Nvidia, AMD, and Intel GPU use and power draw, plus S.M.A.R.T., eMMC wear, and Linux software RAID health. These are the measurements people typically want when a small server becomes slow, hot, full, or unexpectedly noisy.
This focus is also its boundary. Beszel watches machines and containers well, but it is not an application tracing system, log search engine, or general metric query platform. An open feature request for failed systemd service alerts makes the distinction concrete. If a DHCP process dies while the host itself looks healthy, Beszel does not currently offer the service-level check that requester needed.
A setup that stays understandable
The official quick start has five conceptual steps: start the hub, create an administrator, define a system, start its agent, and finish the connection. The hub and agent can run as binaries or containers. For a local Docker host, the documentation provides a Compose file containing both components and a Unix socket connection. A Linux binary installer can create a dedicated user and service, start the agent, and optionally enable daily updates.
That is genuinely easier than assembling Prometheus exporters, storage, alert rules, and a Grafana dashboard. It is not one-click across a fleet. Every remote host needs an agent, credentials or a universal token, a reachable route to the hub, and the correct access to container or hardware information. Non-Linux systems may need manual binary handling. The current pfSense report is a useful warning: one FreeBSD installation did not restart the agent after a router reboot, even though the service configuration remained present.
The hub needs ordinary production care too. Persistent storage is mandatory if history matters. Internet-facing deployments need TLS, a reverse proxy or another safe exposure method, and a deliberate authentication configuration. Beszel supports multiple users, system sharing, OAuth2 or OIDC providers, and disabling password login. Automatic backups can target disk or S3-compatible storage. Those features make it suitable for more than a private experiment, but someone still owns restore tests, secret rotation, and upgrades.
The dashboard is opinionated in the right places
Beszel's strongest product decision is to provide a finished experience instead of a blank graph editor. Systems have consistent charts, containers get resource history, and alerts cover CPU, memory, disk, bandwidth, temperature, load average, and availability. You can copy alerts between systems, a feature added around the latest release, which reduces repetitive configuration as a fleet grows.
The tradeoff is less freedom at the edges. A current request asks for excluding selected containers from alerts because an expected CPU-heavy HandBrake job can create noise. Another open request asks for systemd service failures. Prometheus users could model both cases with custom labels and rules, while Beszel users wait for the product to expose the needed control. That is the bargain: much less assembly, but the application's supported concepts define what is easy.
Mobile login also deserves testing before an operator treats a phone as the emergency console. An open Android report, active since July 2025, describes password-manager autofill preventing the login button from completing authentication across multiple browsers. A recent pull request proposes a fix, but the issue remains open. OAuth may avoid that exact path, yet teams should test their actual identity provider and browser combination.
Active code, busy support surface
The repository was pushed on August 9, 2026, and issues and pull requests were still being updated on August 10. The latest tagged release, 0.18.7, arrived on April 5 and included expanded disk I/O metrics, copied alert settings, macOS ARM64 crash fixes, storage reporting corrections, security-related token restrictions, and dependency updates. The gap since that tag is worth noticing, but current commits and pull-request traffic do not suggest an abandoned project.
GitHub reports 375 open issues and pull requests combined. That is a substantial queue, and the README candidly says the maintainer may not always have time to respond. It also reflects a project with more than 24,000 stars, many platforms, and a wide hardware surface. Recent discussions include GPU detection, FreeBSD startup, container alert controls, and interface fixes. Expect compatibility questions that only appear on specific kernels, drivers, filesystems, browsers, and appliances.
Documentation is a strength. The separate site covers hub and agent installation, reverse proxies, security, OAuth, backups, notifications, extra disks, GPU monitoring, S.M.A.R.T. data, Podman, and common problems. English documentation is the primary path, with a Chinese localization and community translation work visible as well.
Who should choose it
Beszel is a strong first choice for someone who wants historical server and container visibility but does not want monitoring to become its own engineering project. It is especially good for a home lab, a collection of small servers, or a team that values a shared, readable dashboard over custom queries. The MIT license, authentication options, backups, and multi-user support give it room to grow with that environment.
Choose Netdata when diagnostic depth matters more than simplicity, Glances for quick machine-level inspection, Uptime Kuma for endpoint availability, or Prometheus when custom application metrics and alert logic are non-negotiable. Beszel wins when you want the common host questions answered quickly and accept its boundaries. Test agent persistence, mobile authentication, and alert noise on your actual systems before making it the only signal you trust.