Deployport

You didn't get into DevOps for this:

Stop being the help desk for dev VMs.

You have real work to do. With Deployport, developers start, stop, resize, and wake their own machines, within guardrails you define and with every action logged.

Ready for your existing AWS and Google Cloud resources. No migration, no rebuilding instances. Just a trust role between your account and the Deployport account.

One command, on a server you already run.

Free for your first 3 hosts. No credit card.

sudo deployport uplink local install --key "<key>" --region "<region>"

Who it's for

For fast-moving engineering teams only.

Founders

You need to save money and you want your DevOps back.

DevOps

Often the whole platform function is one or two of you. The dev-environment tickets stop landing on your desk, because the policy does the babysitting and the team serves itself.

Developers

Your box stays yours. Wake it yourself, hold it awake through a long build, and nobody has to approve it.

How it works

Our agent daemon, plus our globally available control plane.

Nothing is overkill when it comes to saving you time and money.

us-west-2 us-east-2 Deployport control plane Your cloud Your datacenter

Run your numbers

Pays for itself three times: cost reduction, DevEx and productivity.

Solves a cost issue and a productivity issue in one shot.

Your team
Cost / VM / month

About $300/month. Rough figures, type your own if you know them.

Fully idle days per week (weekends)

Weekends, mostly. Nothing runs, nobody notices.

Per developer, per month
0 hrs of platform engineering
recovered, monthly
$0 a month saved
on idle cloud spend
Show the costs behind this
$0 a month is what these machines cost you today
$0 a month is what Uplink costs, at $7 per host

Wardens

A bunch of mini-yous, so your developers self-serve.

A warden is a pre-built, configurable automation you activate on your machines. Each one has a single narrow responsibility working for you 24/7 to enable self-serve.

  • Alarm

    Notice when a machine disconnects

    Reports a machine that stops responding, and stays quiet when it was stopped on purpose.

  • Alarm

    Report kernel out-of-memory kills

    Names the process the kernel killed, within seconds, from the host's own kernel log.

  • Alarm

    Warn on sustained memory pressure

    Warns when memory use stays high for a whole window, so a build spike never pages anyone.

  • Alarm

    Warn before a disk fills

    Warns on a used-space threshold, and predicts when a volume will fill from its own growth rate.

  • Automation

    Stop machines that are idle

    Turns a machine off once it has been idle, after warning whoever is on it, and lets a developer hold it awake.

  • Automation

    Reclaim disk space with Docker

    Prunes Docker when the disk is under pressure, with a cooldown so a build cache is not killed by a thousand cuts.

  • Automation

    Reclaim disk space from developer caches

    Measures the caches Go, npm, pip and Cargo leave behind, and clears them when you ask. Never on its own.

Questions

The things people ask before they install it.

Do we have to migrate all VMs to Deployport?

No. Deployport works with existing Linux virtual machines, you install the agent and optionally connect your AWS or GCP account with ours for more capabilities.

How is this different from a CDE like Coder or Codespaces?

With Deployport, your VMs are as persistent as you need, and nothing is enforced upon the developers. There is also zero-migration cost to adopt Deployport.

What access to my cloud does this need?

A trust role you create in your own account and can revoke at any moment. You never hand over a key. It also starts at zero: diagnostics need no cloud permissions at all, and each capability asks for its own small set only when you switch that capability on.

What stops it turning off a machine somebody is using?

Idleness is measured. The warden reads real signals from the machine, SSH sessions, logged-in terminals, CPU usage and network traffic, and you set the thresholds. A machine that is working does not qualify, whatever the clock says. Before anything happens the person using it is warned, and they can keep it awake themselves from the desktop app on macOS and Windows, without asking you.

Can I run a long ad-hoc process on my VM?

Yes. A machine that is doing work does not read as idle, so a busy process keeps it up on its own. If the job is quiet enough to look idle and nobody is connected over SSH, you keep the machine awake with one click from the desktop app or the console, for as long as your own policy allows.

Hang on, I could vibe-code this in 15 minutes.

Sure. Version one is a cron job and it really does take 15 minutes. Version two shows up at 4pm the day it kills somebody's build. A quarter later you maintain a product nobody asked for, it has a pager, and it costs you more than we do. Ours is $7 per host, and the pager is ours.

From the founder

We were spending about 9k a month on dev environments, and my one devops engineer was constantly getting distracted by vm-related tickets. At some point we introduced a weekend and nightly schedule to turn those machines off, and it was so inflexible it hurt more than it helped. So I built the version I actually wanted, one that frees up the devops person without the rest of the team feeling it.

Johan, Deployport

The last VM ticket you'll ever close.

curl -fsSL https://get.deployport.com | sh

Your cloud credentials never leave your account. You grant a scoped role, and you can revoke it in one click.