Last month somebody on your team spun up a GPU box "just to test something."

It's still running. Nobody remembers why. The person who started it has changed teams, and the ticket that explains it was closed in a cleanup sprint. Meanwhile it quietly adds a line to your invoice every hour.

Multiply that by every account, every region, and every engineer who has ever typed aws ec2 run-instances on a Friday afternoon. That is your cloud graveyard: unattached volumes, idle load balancers, old snapshots, orphaned IPs, forgotten test stacks.

Everyone knows it exists. Almost nobody cleans it, for three reasons:

  1. Finding it is tedious. The resources are spread over a dozen regions and several accounts.
  2. Deleting is scary. "What if something still depends on it?" is a fair question, and nobody wants to answer it at 2 a.m.
  3. Nobody owns it. If you can't say whose it is, you can't ask them to delete it.

So I built Recoup, a tool aimed at those three problems.

What Recoup does

It finds the stale stuff, across accounts and regions.
Point it at your AWS accounts with read-only credentials. It sweeps the regions and lists resources that look forgotten.

It tells you whose they are, and which code made them.
Each finding is attributed to an owner and cross-referenced to the code that created it. "Who do I ask?" stops being a Slack archaeology project.

It shows its maths.
Every monthly-waste estimate comes from a formula you can inspect. If a number looks wrong, you can see exactly why it is that number, and argue with it.

It writes the cleanup, but never runs it.
You get tiered cleanup scripts, from safe to risky, and they are reversible. Nothing is executed from the browser. You read the script and you run it, or you don't.

It remembers.
The History page asks the uncomfortable question: did the waste come back? Cleaning up once is easy. Staying clean is the hard part, and recurrence shows you whether the same leak reopened.

It makes the case to your finance team.
There's a savings report as a PDF, built for the person who approves the budget and doesn't care what a NAT gateway is.

The part I care most about: it's read-only

A tool that scans your cloud account has to earn trust before it gets any. So:

  • It needs read-only credentials.
  • It doesn't delete anything.
  • Nothing runs from the browser.
  • You can revoke its access at any time.

It hasn't had an independent security review yet, and I won't claim it has. That is planned. For now, the safest design I could think of is one that can't change your account at all.

Who it's for

  • Platform and DevOps engineers who inherited an account they didn't build
  • Startups whose AWS bill grew faster than their revenue
  • FinOps folks who are tired of sending spreadsheets that nobody acts on
  • Anyone who has said "we should really clean that up" more than twice

It's a pilot, so it's invite-only

Recoup is in a closed pilot. Sign-up needs an invite code, and I'm handing them out in small batches so I can talk to every early user.

To request one:

📧 Email hello@recoupify.in with the subject Recoup invite.
Tell me in a line or two how many AWS accounts you run and what hurts the most. I read every message.

🌐 Site: https://recoupify.in

Early users get a direct line to me, and their feedback decides what gets built next.

One question for the comments

What's the most expensive thing you ever found running in AWS that nobody could explain?

Mine is a story for another post. I'd like to hear yours first.

If this was useful, a ❤️ or a 🦄 helps other engineers find it. If your AWS bill is a mystery, you know where to email.