Planned · being built now

Keep an eye on
your AI machines.

Running AI models or agents on your own graphics cards? ServerNest will show how every card is doing, right next to the rest of the server, in the same simple dashboard. You'll see when a card runs hot, when its memory is full and which program is using it.

This is planned and being built now. It is not in the preview yet.

Join early access What you'll see
  • NVIDIA cards first
  • Same small agent
  • No open ports
  • Your machine, your data
Illustration of the planned view. The numbers are examples, not a real server.

What you'll see

Everything that matters for AI work.

Each graphics card gets its own card on the server's page, with small graphs over time. Servers without a graphics card stay the same.

How busy each card is

See how hard every graphics card is working, now and over the last hours.

Video memory

How much of the card's own memory is in use. So you know if a bigger model will still fit.

Heat and fan

How hot each card runs and how fast its fan spins, before heat becomes a problem.

Power and slow-downs

How much power a card uses compared with its limit, and when it slows itself down to stay cool.

Which programs use it

Every program on the card and how much memory it holds. See at a glance which model or agent is busy.

Network and disk speed

Download and upload speed, and how fast the disks read and write. Handy when you pull a large model.

Plus what ServerNest already shows for every server in the preview: CPU, memory, disk space, uptime and whether a restart is needed. How monitoring works today.

Alerts

Know before a long job fails.

We plan alerts for the things that stop AI work: a card that runs too hot, video memory that is full, a card that slows down, or a card the server can no longer see.

They'll go where your other ServerNest alerts already go: email, the bell in the dashboard, Slack, Discord, Telegram or your own webhook. Each person picks which alerts they get.

Illustration. GPU alerts are planned and not built yet.

Who it's for

For people who run AI on their own hardware.

Local AI models

You run models on your own graphics cards, with tools like Ollama or vLLM, and want to see what the card is doing without logging in.

AI agents that run all day

Your agents work on a server around the clock. You want to know when something gets stuck, full or hot.

Rented or home GPU servers

A GPU server from a hosting company, or a machine under your desk. It needs no open ports, so both work the same.

Where it stands

Honest status.

We only call something ready when it's tested. Here is where AI machine monitoring is today.

  1. In development

    The basics

    Reading NVIDIA cards, network speed and disk speed, with graphs on the server's page.

  2. Planned

    Alerts and lists

    Alerts for heat, full memory and missing cards, and graphics card details in the server list.

  3. Coming later

    More cards

    AMD graphics cards, and a one-time speed test for network and disk that you start yourself.

Which plans will include it is not decided yet. During the private beta everything is free. Have an AI server you'd like us to support? Tell us what you run.

Questions

Does it need extra software on the server?

No. It uses the same small ServerNest agent. For NVIDIA cards it reads the numbers the NVIDIA driver already gives, so the driver must be installed.

Can ServerNest change settings on my graphics card?

No. It only reads numbers. It doesn't change clocks, power limits or fans, and it can't see what your models are working on.

Will it slow down my models?

It is designed to stay out of the way: the agent reads the numbers about once a minute and sends a small message. Servers without a graphics card send nothing extra.

Can I run websites on the same server?

Yes. ServerNest already manages websites, PHP and databases. An AI server is still a Linux server, so you see both in one place. See supported systems.

Want to try it first?

Early-access users try new features before anyone else, and help decide what we build next.

Join early access See all features