You write queries. We do the rest.

At the moment, you can only run PostgreSQL.

Three things, and nothing else: how many servers you run, the compute size of each, and the storage ceiling attached to it. Compute is a rung from the ladder — vCPU and memory sold together, from 1 vCPU · 1 GB up to 64 vCPU · 512 GB. Storage is however many 10 GB blocks you want, at €3 each. The databases you create on a server are not counted or charged, and data transfer is not billed.

No, and that is on purpose: the ladder is priced in a straight line, so a rung with twice the cores and twice the memory of another costs exactly twice as much. There is no tier where the rate improves and none where it worsens, so picking a size is a question of what your database needs rather than where the pricing happens to bend.

No. Every 10 GB block is €3 — the first one and the hundredth — which works out at €0.30 a GB at any size. There is no allowance to use up first and no tier where the rate changes, so the storage line is one multiplication however large the server gets.

Twice the price of that server. The standby runs at the same size and keeps its own full copy of the data, so the compute line and the storage line both double — there is no half-price second copy. Turn HA on or off whenever you like, and the change is prorated.

Point-in-time restore is included at every server size and whatever storage ceiling you set — no tier adds it and there is nothing to buy. Scheduled full backups are the one optional line: pick daily, weekly or monthly on all the databases of a server or only some of them, and buy a backup storage quota at €1 per 10 GB a month to hold them. The two do different jobs — PITR rewinds a server inside 15 days, a backup is a copy you still have months later.

By default, no. Servers answer on the private network the services in a project share, so your API reaches yours without crossing the internet. Public access is opt-in and restricted to the addresses you list.

As many as the hardware will carry. It is a plain PostgreSQL server, so you create databases on it the way you would anywhere else, and we neither count them nor charge for them. What a given size holds comfortably depends on your connections, working set and write rate, so the dashboard offers guidance rather than a cap.

pgvector, PostGIS, pg_stat_statements, uuid-ossp and the rest of the common set, enabled per database rather than per plan.

Either one, whenever. Changing a server size, a server count or the storage ceiling takes effect immediately and is prorated from the moment you make the change, so growing is a setting rather than a migration.