Deploy your first app

Prisma Compute is serverless hosting for TypeScript apps. It runs your app right next to your Prisma Postgres database, so the trip from app to data stays short.

This is a quickstart: sign in, then deploy. There are two ways in:

• To start fresh, scaffold a new app and deploy it in one step with create-prisma. Follow Option A.
• To deploy an app you already have, run a deploy from its directory. Follow Option B.

To learn more about how Compute works, see the resource model for a mental model.

Prerequisites

Before you start, make sure you have:

• A JavaScript runtime. Node.js 22.12 or newer for npx/pnpm dlx, or Bun for bunx.
• A Prisma Data Platform account. Free to create, and it holds your workspace.
• An app, if you are bringing your own. Compute deploys Next.js, Nuxt, Astro, Hono, TanStack Start, and plain Bun servers today, including Elysia (which runs on Bun). Skip this if you are scaffolding a new app below.

export default { output: "standalone" };

1. Sign in

Sign in once. This is the only step that needs a human, because it opens your browser:

bunx @prisma/cli@latest auth login

It stores a session on your machine that every later command reads automatically, including a coding agent working in your directory. Confirm it any time with auth whoami.

Two shortcuts from here:

• In CI or a headless environment where no browser can open, use a service token instead.
• To let an agent do the deploying for you, jump to Hand it to your agent.

2. Deploy your app

Both options finish in the same place: a live URL and a project you can deploy to again.

Option A: Scaffold a new app with create-prisma

create-prisma builds a TypeScript app from a template, wires up Prisma, and can deploy it to Compute in the same run. Run it and answer the prompts:

bunx create-prisma@latest

It asks for a template, a database provider, and a few options, then asks "Deploy to Prisma Compute now?". Say yes, and the deploy step:

• signs you in, if you have not already
• creates a Prisma project for the app
• builds and deploys it, then prints the live URL
• provisions a Prisma Postgres database and wires it in, when you pick PostgreSQL and don't bring your own connection string

So your deployed app has a database from its first request.

To skip the prompts, pass your choices as flags. Compute deploys are available for the hono, elysia, next, and tanstack-start templates:

bunx create-prisma@latest --name my-api --template hono --provider postgresql --deploy

When it finishes, it prints your live URL and adds a compute:deploy script so you can redeploy later. Then verify the deployment.

Option B: Deploy a project you already have

From your app's directory, run:

bunx @prisma/cli@latest app deploy

In one pass, the CLI:

1. Detects your framework from your project files, whether that is Next.js, Nuxt, Astro, Hono, TanStack Start, or a plain Bun server. To choose it yourself, pass --framework (use --framework bun for a Bun or Elysia server).
2. Sets up a project the first time you deploy from this directory, then writes .prisma/local.json to pin the directory to that project. That file is a gitignored local cache, not committed config. If your team already has a project, link it first.
3. Resolves the target branch. Inside a Git repository, the CLI uses your current Git branch name; otherwise it falls back to main. Pass --branch <name> to choose explicitly. Because each branch is its own isolated environment, this decides where the deploy lands.
4. Builds and uploads your app, provisions it, and prints a live URL.

Your first deploy is promoted to production automatically, so the URL you get back is your production URL.

Need a database too? Add --db, and the deploy provisions a Prisma Postgres database and wires its connection string into this deploy target:

bunx @prisma/cli@latest app deploy --db

3. Verify the deployment

You have a URL. Confirm the app responds by opening it:

bunx @prisma/cli@latest app open

If it does not behave the way you expect, stream its logs:

bunx @prisma/cli@latest app logs

Logs cover both the build and the running app, so a failed build and a runtime error both surface in the same place.

You can also inspect everything you have deployed in the Console instead of the terminal:

• projects, branches, and apps
• deployments, integrations, and domains

4. Deploy again

Run app deploy whenever you want to ship a change. After your first deploy, production is protected: deploying to your production branch again needs an explicit --prod flag, so you never ship to production by accident.

bunx @prisma/cli@latest app deploy --prod

Deploying from a Git feature branch behaves differently. Instead of touching production, the CLI:

• resolves the branch name and creates a matching platform branch if one does not exist yet
• gives you an isolated preview deployment with its own app, database, and URL

Preview deploys never ask for confirmation, so run them as often as you like. In scripts and CI, add --yes to accept the production confirmation up front:

bunx @prisma/cli@latest app deploy --prod --yes

5. Set up deploy on push (optional)

Connect your project to a GitHub repository once, and Prisma deploys every push for you:

• every commit you push is built and deployed on its own
• each branch gets its own preview, with its own URL
• you stop running app deploy by hand

Make sure your project directory is a Git repository with a GitHub remote, then connect it:

bunx @prisma/cli@latest git connect

git connect wires up the project this directory is linked to, which exists once you have deployed at least once. For monorepos and CI, see the GitHub integration docs.

Hand it to your agent

You can let a coding agent do the deploying. Sign in once yourself (step 1), because the browser step needs a human. After that, anything running in your environment inherits the session, including your agent. Paste this into your agent and fill in the blanks:

Build [what you want] in [framework] and deploy it to Prisma Compute using `npx @prisma/cli@latest`.

For example:

Build a Hono API with a /todos endpoint backed by an in-memory list. Deploy it to Prisma Compute using `npx @prisma/cli@latest`.

Notes for your agent:

• Run every CLI command as npx @prisma/cli@latest <command>, and add --json for structured output.
• Check login state with npx @prisma/cli@latest auth whoami.
• On the first deploy, the CLI creates the project and prints a live URL. Open it and confirm the app responds. The first deploy is promoted to production automatically.
• To give the app a database, add --db to the deploy so Compute provisions and wires a Prisma Postgres database.
• If the app needs config or secrets, scope them to the environment you're deploying. Use --role production for production deploys and --role preview for preview branches, then redeploy: npx @prisma/cli@latest project env add KEY=value --role production (or --role preview). Don't write production config to the preview scope. For the full scoping rules, see Environment variables.

What's next

• Branching: how preview branches isolate work and map to your Git branches.
• Add environment variables for configuration, secrets, and your database connection string.
• Deployments: promote, roll back, and inspect what you ship.
• GitHub integration: the full picture on deploy-on-push, monorepos, and cleanup.
• Read the full CLI getting-started guide for frameworks, project linking, and CI.