Databases

A managed database is billed by the hour it exists, not the hour it is queried.

That single sentence is why a small app costs fifty dollars a month somewhere else and cents here. Everything on SproutOS suspends when nothing is happening, and for almost every app, almost nothing is happening almost all of the time.

Postgres
Your data. A real Postgres you can connect to with a connection string, not a wrapper with a query language of its own.
Suspends when idle and wakes on the next connection.
Valkey
Caching and queues. Tenant-split, with an engine-enforced identity rather than a prefix convention everyone has to remember.
Shared capacity, so an idle queue occupies almost nothing.
OpenSearch
Search, when a LIKE query stops being enough. Also tenant-split, and also not something you have to name or size in advance.
Storage while it sits there; you pay for queries when they happen.

You do not need to know which of these you want. You describe what the app should do, and SproutOS picks the pieces and wires them together.

Priced both ways

The same three things, rented yourself.

These are the cheapest options each vendor sells — the smallest instance and the smallest disk — not a typical bill. The gap is not a discount we are funding. It is that none of them can be turned down when idle.

Databases

Postgres, a cache, and search

SproutOS gives you all three: Postgres for your data, Valkey for caching and queues, and OpenSearch for search. Each is tenant-split and suspends when nobody is using it, which is the entire reason the number below is a number of cents.

Doing it yourself

RDS Postgresdb.t4g.micro + 20 GB — the smallest one AWS sells$13.98
ElastiCachecache.t4g.micro$11.68
OpenSearchone t3.small.search node$26.28

All three, always on$51.94

On SproutOS

$0.02/mo

1,000 accounts

On Supabase's free tier, two is the most databases you can have running at once.

None of these can be turned down when idle. That is the same $51.94 whether a thousand people use your app this month or nobody does.

Why a database per person is possible at all

At fifty dollars a month per stack, one database per user is a business model nobody can run. At cents, it is just how you would build it if you were not worried about the bill — which is why the data-ownership argument on this site is a consequence of the pricing rather than a slogan attached to it.

What that gets you

Comparison figures are published list prices for us-east-1, read from the AWS pricing API in August 2026 and multiplied by 730 hours, and are the cheapest option each vendor offers rather than a typical bill.

How it is billed

Usage, prepaid, with no monthly floor.

There is no seat, no plan and no minimum spend — you top up credit and work draws it down. Storage is charged per gigabyte-month, compute per second it is awake, and queries per query. When credit runs out, new work is refused rather than quietly billed to a card.

The exact per-dimension rates and the platform fee on each are in the billing documentation.

Get started