Why every project gets its own Postgres

Most AI builders keep every customer's data in one shared database, partitioned by a tenant column. It is the cheapest architecture to operate, and it is exactly what a founder, an agency or an IT team does not want to explain to their own customers.

What a cloud project runs on

When you provision a cloud project here, the Builder creates a dedicated server for it (or adds it to one of your existing servers, if you choose; Europe on ARM or the United States on x86), installs PostgreSQL 16 on that server, and runs the app as a systemd service behind Caddy. The database is yours in the plainest sense: it is on a machine that runs only your projects.

What that buys you

• No noisy neighbours and no shared blast radius. A bug, a migration or an agent rewrite in someone else's project cannot touch your rows.
• Ordinary tooling. pg_dump, psql, connection strings, extensions: it is a normal PostgreSQL 16 install.
• A clean exit. If you self-host later, you dump and restore a database; you do not extract your rows from a shared one.

What it costs

Dedicated servers are on paid plans and are billed as a flat monthly price per server size. The real numbers, read from the settings the product bills from, are on the pricing page.