Your servers and your network, on one screen.

Proxmox, in depth

proxmon reads each host through the Proxmox API every 30 seconds. Nothing is installed on your hosts to monitor them. Sites and groups you arrange hold your hosts down the left, each with its VMs and containers under it.

Hosts

CPU and memory with 24-hour graphs (now, average and peak), root disk, uptime, Proxmox version, API latency, and whether the host or only its API stopped answering.

VMs and containers

Status, CPU and memory with small graphs on every row, disk use from inside the guest, IP addresses from the guest agent, uptime, and the hardware from each guest's config.

Storage, disks and ZFS

How full every storage is and when it will fill, SMART health with each drive's temperature, hours, SSD life left and error counters, the full SMART report, and ZFS pool state.

Hardware health

Temperatures, fans, power supplies, power draw, memory errors and the server's hardware event log, from a small read-only helper set up with one pasted command. Reads Dell, HPE, Supermicro and IBM management controllers without their password.

Tests and burn-in

SMART short and long self-tests, a read-only scan of every sector, and CPU and memory burn-in from 15 minutes to 24 hours, with live progress, a Stop button and an automatic stop near a CPU's critical temperature.

Tasks and logs

Backups, starts, stops and migrations with their logs, and each host's live system log with a filter.

Consoles and power

On hosts you give full control: a VM's screen and serial port, a container's console and the host's shell, each in a browser tab, plus start, shut down, restart and force stop. Anything that interrupts a running guest asks first, and every action and console session is logged.

Your network, as a NOC

Switches, routers, firewalls and anything else that speaks SNMP (v1, v2c or v3) sit in their own section beside Proxmox. Add a device, test its credentials, tick what to watch, and proxmon polls it every minute: every port's traffic, utilisation, errors and link state, and CPU, memory and storage where the device reports them.

Ports three ways

Rows with live traffic and graphs, a grid of boxes for 48-port switches, or a plain list. Each port is named from the description set on the switch, and opens to its own traffic, utilisation, error and discard graphs.

Yellow before red

A missed poll or a dropped link is yellow first and turns red after the device's "down after" count, so one lost packet doesn't page anyone. Ports you don't expect up stay grey.

Levels where you want them

Warning and critical levels on a port's utilisation or a device's CPU, memory or storage, with a delay ("only after 5 minutes") so a burst doesn't page anyone.

Polling that's kind to your network

Every device polls at its own second, so a hundred devices added together don't all ask at once, and an optional per-network limit keeps a rate-limited router happy without ever counting the wait as a miss.

Choosing what to watch

Discovery lists everything a device reports and ticks what's up. Come back any time to stop watching unused ports (they stop counting toward your plan and leave the device's page), turn off "alert when down" for a port that's only sometimes used, or rename, with buttons for "ports that are up" and "ports that are down".

Vendors

Anything with standard SNMP interface counters works: Juniper, Cisco, Arista, MikroTik, HPE and Aruba, Fortinet, Palo Alto, Ubiquiti, Synology and more, with CPU, memory and storage where the device publishes them in the standard host-resources table. Where SNMP leaves gaps, a vendor integration fills them.

• UniFi. With a read-only API key, ports get their faceplate number and the name you gave them in UniFi ("Port 10 – Juniper EX switch"), link state comes from UniFi (whose SNMP reports every built-in port as up), and VLAN, bridge and VPN interfaces are labelled with your network names ("br10 – Cameras (VLAN 10)").
• Import from PRTG. Upload PRTG's device export, or any CSV with an address column. A preview shows what's new and what's already here before anything is added, PRTG's groups become folders, and each device picks up its sensors the first time it answers.

Alerts that reach you

An alert opens while a problem lasts and clears on its own when it's fixed. Each one goes to the channels you choose, and nobody gets an alert about something their role doesn't cover.

What it catches

A host, guest, device or link down; storage and disks filling; failing drives, worn SSDs and rising error counters; hot drives; degraded ZFS pools; failed backups and tasks; ports over their levels; hardware sensors out of range; websites that stop working.

Where it goes

Per person: ntfy topics, email, webhooks (Discord, Slack or plain JSON), and push to proxmon installed on a phone or desktop. Each channel picks its severities and quiet hours, with criticals allowed through.

Calm when it should be

Delays so a spike doesn't page anyone, maintenance windows and pauses for a group, host or device, and "mark as seen" to grey out what's being handled until it gets worse.

Who watches the watcher

A dead-man's switch: proxmon pings healthchecks.io, Uptime Kuma or another proxmon after every poll, and two proxmon installs at two sites can watch each other.

Service checks

A website that answers "200 OK" with the wrong page is down to your customers. A service check asks for a URL and checks the status, the text the page must contain (or a regular expression) and how long it took, from proxmon's place in your network. Attach one to the guest that serves the site and it shows on that guest's page.

Reports

Health and usage over any period, in the browser or as a PDF, for all hosts or one, up to 13 months back.

Health

What needs attention, drives and how their SMART counters changed, hardware sensors, tests run, backups and alerts.

Usage

CPU and memory average, 95th percentile and peak, availability and outages, guests, storage growth and when it fills, and website uptime.

On a schedule

Weekly or monthly, emailed to whoever needs it.

Built for a team

One proxmon for the whole NOC, with each person seeing what's theirs.

Roles by part of the tree

View, operate, edit or admin, for the groups, sites, hosts or devices a role covers. Anything outside someone's roles is simply absent for them: sidebar, overviews, alerts, reports and notifications.

Directory sign-in

Active Directory, OpenLDAP or FreeIPA over LDAPS or StartTLS, with directory groups mapped to roles at every sign-in.

Two-step sign-in for everyone

A password plus an authenticator app code on every account, and an activity log of every sign-in and change.

Live, on any screen

Dashboards update in place without reloading, for a wall screen in the NOC. Installed on a phone, proxmon is an app with its own notifications.

Across your sites

Your hosts and devices don't have to share a building. Servers at head office, a rack at a branch, a cage in a colo: proxmon watches them all from one page, over a private encrypted network between your sites, with nothing opened to the internet.

No port forwarding

Hosts and devices are reached at their private addresses, over the encrypted network or directly when proxmon has a leg in each network. Nothing has to face the internet.

Each host proves who it is

proxmon pins every host's certificate when you add it, so a machine pretending to be your host is refused, even inside the network.

Least access

Each Proxmox host gets its own token, read-only by default. SNMP credentials are stored encrypted, and changing them means testing them first.

Your network decides

With Tailscale, your own access rules say what proxmon can reach, and you can cut it off from your side at any time.

Hosted proxmon Planned

Rather not run it yourself? We'll run your proxmon for you, in our data center with geographically diverse connections, and connect it into your tailnet.

• One instance per customer. Your proxmon runs alone in its own sealed-off network segment. It can't see other customers, and they can't see it.
• Nothing open to the internet. You reach your proxmon through your own tailnet. It accepts nothing from the public internet, and its outgoing traffic is limited to what it needs to work.
• You hold the keys. Your Tailscale rules decide which hosts and devices it may reach. Remove it from your tailnet and it's cut off, immediately.
• Read-only unless you say otherwise. Control features stay off until you turn them on per host, and every action is in your activity log.

Getting started

Once your access is approved you get a license key and one command. proxmon installs on a small Debian VM or container (1 CPU, 1 GB of memory is plenty) or in Docker, and updates itself with one more.

1. Install

   Paste the command from your approval email into a fresh Debian 12 or 13 machine. Add Tailscale in the same step if your hosts are at other sites.

2. Add your Proxmox hosts

   Type the address you use for the Proxmox web page. proxmon shows the host's certificate for you to confirm, creates its own limited token from the root password (which it doesn't keep), and refuses a token with more access than it needs.

3. Add your network

   Add a switch, router or firewall with its SNMP credentials, test them, and tick what to watch. Or import the lot from PRTG or a spreadsheet.

4. Tell it where to reach you

   Add your phone, email, or a Discord or Slack channel under notifications, and arrange the tree into the sites and groups you think in.

Security

A tool that can see every server and switch you run should be careful with that.

• It runs where you choose. On your own network, where the only thing it sends us is a daily update check with your license key and version number, or hosted by us in an isolated instance that only your tailnet can reach.
• Read-only by default. A new host's token can look and can't change anything. Power controls and consoles are a separate step you take per host, and network devices are only ever read.
• Two-step sign-in for everyone. Every account needs a password and an authenticator app code. Password reset by email still needs the code.
• Pinned certificates. Each host's certificate, and a vendor console's, is remembered, so a different machine at the same address is refused.
• Encrypted secrets. API tokens, SNMP credentials, vendor API keys and authenticator keys are encrypted in proxmon's database, and every sign-in and change is in an activity log.
• Verified updates. Each release is checked against its SHA-256 before it installs, and a release that fails to install puts the previous one back.

Roadmap

proxmon is in early access. Everything above works today; the rest comes one careful step at a time.

1. Proxmox monitoring

   Hosts, guests, storage, disks, hardware health, tests, tasks, logs, phone alerts, the add-host wizard.

2. Control

   Start, shut down, restart and force stop guests; VM, container and host consoles in the browser.

3. Alerting and reports

   Thresholds, notification channels, quiet hours, maintenance windows, service checks, a dead-man's switch, PDF and scheduled reports.

4. Teams

   Roles by part of the tree, directory sign-in, the activity log.

5. Network monitoring

   SNMP v1, v2c and v3 devices, the Network Overview, port grids, Manage sensors, UniFi, import from PRTG, staggered polling.

6. Deeper network health

   Temperatures, fans and power supplies on switches and routers; built-in CPU and memory for Cisco and Juniper; PRTG's sensors in the import; very large routers.

7. Snapshots and backups

   Take and roll back snapshots, back up a guest now, restore.

8. Changes

   Adjust CPU, memory and disks, create guests from templates, move guests between hosts.

9. Hosted proxmon and remote collectors

   Your own isolated proxmon, run by us and connected to your tailnet. How it's kept separate. And small collectors for networks proxmon can't reach directly.

Pricing

One meter for everything: a sensor is one thing watched, such as a switch port's traffic, a server's CPU, a VM, a disk or a website check. A server's hardware health counts once, however many temperatures, fans and power supplies its management controller reports. Every plan has every feature; you choose how many sensors you need.

Prices in US dollars, before tax, billed yearly. Counting works the way you'd expect from other monitors: one sensor per interface, CPU, memory, disk, guest or check. Online checkout isn't open yet: choose a plan in the request form below and we'll follow up by email about payment. If a paid plan ends, your license continues on the Free plan. Every download is under the End User License Agreement, which you accept once at proxmon.net/accept after your access is approved. The plans cover everything on this page: Proxmox hosts and guests, network devices and their ports, and service checks.