Three databases, all of them ours to run

Create one, link it to the services that should reach it, and the credentials arrive as environment variables on the next deploy. Nothing listens on the public internet, nothing is injected behind your back, and the server is our problem rather than yours.

libsql

Every project can have databases, and each one is a libsql instance we run: the SQLite dialect you already know, spoken over HTTP so there is no connection pool to exhaust and nothing to keep warm between requests. It is the one to reach for when your service sleeps between visitors.

Best for
Services that sleep between visitors, and read more than they write
Speaks
SQLite, so your queries and migrations already work
Branching
Fork production data and rehearse the migration against it
Included
50 MB on Free, 5 GB on Pro
Read about libsql →

Postgres

Postgres when your schema wants Postgres: extensions, window functions, types that libsql does not have, and every tool you already own. You create the database, link it to the services that should reach it, and the connection string arrives as an environment variable. We run the server.

Best for
Schemas that want real Postgres, concurrent writers and analytics
Extensions
PostGIS, pgvector, full text search
Works with
Prisma, Drizzle, psycopg, ActiveRecord, and every psql you own
Included
50 MB on Free, 5 GB on Pro
Read about Postgres →

Valkey

The cache, the queue and the rate limiter that every app eventually wants. Valkey is the open fork of Redis, so it speaks the same protocol and works with the client you already import: the change is one URL. You create it, link it to a service, and we run it.

Best for
Caching, queues, sessions and rate limits
Speaks
The Redis protocol, so your client library does not change
Licence
BSD, under the Linux Foundation, which is why we can run it at all
Included
50 MB on Free, 5 GB on Pro
Read about Valkey →

Go beyond what seems possible.