Prisma Compute
Deploy TypeScript apps alongside Prisma Postgres, with hosting, database branches, and previews managed together.
Prisma Compute is TypeScript app hosting built to run alongside Prisma Postgres. It is the last step of the recommended Prisma stack: you model your data with Prisma 8 or Prisma ORM, store it in Prisma Postgres, and deploy your app with Prisma Compute, so your schema, database, hosting, and branch previews work together in one Prisma project.
Prisma Compute is in Public Beta. You manage it with the unified Prisma CLI (published as @prisma/cli@next until launch), and deployments come from a git push, the Console, or Prisma Composer.
Get started
The best way to understand Compute is to deploy an app. The quickstart takes a TypeScript app to a live URL in a few minutes, then connects GitHub so every push can deploy automatically.
Deploy your first app
The quickstart: sign in, deploy, verify, and connect GitHub.
Start from the beginning
No app yet? Scaffold one with Prisma 8 and Prisma Postgres first.
Working with a coding agent? The deploy quickstart ends with a ready-to-copy prompt in Hand it to your agent.
To learn how Compute works, the model below shows how each branch gets isolated app and database resources, so preview work never touches production.
The model
Compute organizes everything into a few resources:
- A project groups one product or codebase.
- A branch maps to a Git branch in the linked repository, and gets its own isolated app, environment variables, and deployments.
- An app is an HTTP service, such as a frontend or backend, and can have multiple deployments.
The branch is the concept that ties Compute together. Each branch owns its own app, URL, and environment variables, and can have its own database. When you deploy to a new branch, Compute provisions isolated resources for that branch, so you can test changes in preview without affecting production.
- Production: the default Git branch (e.g.,
main) is your production branch, with its own resources and environment variables. - Preview: every other branch is a preview branch, with its own isolated resources and environment variables scoped to preview.
After you connect a GitHub repository, pushing to a Git branch builds and deploys its preview automatically.
The animation below shows how a project, its branches, and their infrastructure fit together, and how each branch's variables resolve by scope. Step through it, or let it play:
For the full picture, see Branching.
CLI and Console
Use the CLI for anything you want to repeat: local development, CI, and agent workflows. Every command supports a --json flag that switches its output to machine-readable JSON, so scripts and agents can parse results instead of scraping terminal text.
Use the Console if you prefer to view your resources instead of running a command. The Console shows your projects, branches, apps, and deployments in a visual way, and lets you manage them with buttons instead of commands.
What to read next
- Getting started: sign in, write the compute config, connect GitHub, and deploy.
- Branching: how branches isolate work and map to Git.
- Deployments: build, deploy, logs, promote, roll back.
- Image Transformations: resize, encode, cache, and serve images from a Compute app with Bun.
- Environment variables: production, preview, and per-branch overrides.
- GitHub integration: connect a repo and deploy on push.
- CLI reference: every command and flag.
- Domains: point a custom domain at production.
- Known limitations: what the beta can and can't do.
- FAQ: quick answers to common questions.