Postgres

A Prisma.Project owns everything; a Prisma.Postgres database lives inside it. Declare both and alchemy converges them like any other resource.

Project

import * as Prisma from "alchemy/Prisma";

const project = yield* Prisma.Project("app", {
  createDatabase: false,
  region: "us-east-1",
});

By default, creating a project also provisions a default Prisma Postgres database in the chosen region. Pass createDatabase: false (as above) when you’d rather declare databases explicitly — the recommended shape, since standalone databases have their own lifecycle and outputs.

Database

Prisma.Postgres and Prisma.Database are the same resource under two names — use whichever reads better:

const postgres = yield* Prisma.Postgres("db", {
  project,
  region: "us-east-1",
});

Omit name and alchemy generates a stable physical name, so an interrupted create recovers instead of duplicating the database.

Branches

A Prisma.Branch groups databases under a git-style name, and a database attaches to one by branchId (or branchGitName):

const branch = yield* Prisma.Branch("preview", { project });

const previewDb = yield* Prisma.Postgres("preview-db", {
  project,
  branchId: branch.branchId,
});

See Branches for preview-per-stage patterns and default-branch promotion.

Local dev

Under bun alchemy dev, the database resolves to a local @prisma/dev Postgres instead of the cloud — same outputs, zero provisioning. The dev prop tunes it:

const postgres = yield* Prisma.Postgres("db", {
  project,
  dev: {
    // run migrations once the local server is ready, with
    // DATABASE_URL pointing at it
    migrate: "bun run db:setup",
  },
});

persistenceMode: "stateless" gives you a fresh database on every dev session; the default ("stateful") keeps data across restarts.

Where next

Connections — credentials, connection strings, and origins for Hyperdrive.
Compute — deploy an app next to the database.

Reference: