Prisma

Prisma’s data platform gives you serverless Postgres and a managed app runtime. With alchemy you declare the project, its databases, the connections your apps use, and the Compute deployments that run them — all as resources in one Stack. In dev mode the same Stack runs against a local @prisma/dev Postgres with nothing to provision.

New here? Set up credentials, then follow the four-part tutorial to deploy a Project, an Effect-native API, Postgres, and a Vite frontend.

Resources

A Prisma.Project is the top-level container. By default it creates a Prisma Postgres database in us-east-1:

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

Prisma injects that database’s system-managed DATABASE_URL and DATABASE_URL_POOLED variables into Compute deployments automatically.

When you need a database with an independent lifecycle and explicit connection outputs, declare it separately:

const project = yield* Prisma.Project("app", { createDatabase: false });

const postgres = yield* Prisma.Postgres("db", {
  project,
});

const connection = yield* Prisma.Connection("api", {
  database: postgres,
});

Display names are optional for these resources — omit them and alchemy generates stable physical names from the Stack, logical ID, and stage.

The connection exposes ready-to-use outputs — a conventional databaseUrl, direct and pooled connection strings, and parsed origin components for poolers like Hyperdrive. See Connections.

Branches group databases and apps under git-style names. A standalone Prisma.Branch has Prisma’s preview role even when promoted to be the project’s default branch.

Websites

Use Prisma.Website for Vite, Astro (SSR or static), Next.js, Nuxt, SvelteKit, React Router, SolidStart, TanStack Start, Waku, Octane, Foldkit, Vocs, or any static build. The framework’s Node-target output runs on Bun in Prisma Compute, uploaded as tar.gz — no Docker image or registry. Even static sites run a Compute static-file server.

Omit project for a database-less project created only on live deploy. alchemy dev runs the native framework dev server without creating Prisma resources for the Website. Start with the Vite guide or framework overview.

Compute

Prisma.Compute builds, uploads, and promotes an application next to its database — either a framework app built from a directory, or an Effect-native app defined inline:

const app = yield* Prisma.Compute("api", {
  project,
  path: "./app",
  build: "auto",
  healthCheck: { path: "/api/health" },
});

For a standalone database, pass connection.databaseUrl explicitly as shown in Connections.

The candidate health check gates promotion. If the stable endpoint fails after promotion, alchemy attempts to restore the previous deployment and fails closed if recovery cannot converge. See Deployments.

Runtime bindings

Inside a Compute app, Lambda function, or Cloudflare Worker, Prisma.Connect turns a Connection into a typed runtime client whose databaseUrl feeds straight into Prisma ORM v8, SQL.Postgres, or Drizzle:

Effect.gen(function* () {
  const db = yield* Prisma.Connect(connection);
  const sql = yield* SQL.Postgres({ url: db.databaseUrl });

  return {
    fetch: Effect.gen(function* () {
      const users = yield* sql`SELECT * FROM users`;
      return yield* HttpServerResponse.json(users);
    }),
  };
}).pipe(Effect.provide(Prisma.ConnectBinding));

The Prisma path

On Cloudflare, Prisma Postgres slots into the same path as the other Postgres providers:

Hyperdrive pools the connection’s direct origin at the edge
Prisma ORM v8, Drizzle, or SQL.Postgres gives the Worker a typed query layer over that connection
Connect from Workers walks through both wiring styles