Prisma Next

Prisma Next is a full TypeScript rewrite of Prisma ORM. It keeps the schema-first workflow and model-first queries you already know, and rebuilds everything underneath to be extensible, composable, and AI-agent friendly by default.

Instead of compiling your schema into a heavy generated client, Prisma Next treats your schema as a versioned contract: open, inspectable artifacts that your code, your tools, and AI agents can all rely on.

Why Prisma Next

Prisma's current architecture tightly couples three layers: the schema language, the generated client, and runtime execution. That coupling makes Prisma harder to extend, slower to rebuild, and opaque to the tools that read your code. Prisma Next rebuilds the data layer around a contract-first model with a small, extensible core:

• Better queries: a cleaner API with simpler nested queries, custom collection methods on your models, and streaming results, plus a low-level, type-safe SQL query builder for the cases that need raw control.
• Extensible by design: a minimal core exposed through a public SPI. Everything around it, including Postgres support itself, is an extension, so you can add databases, query builders, data types, and middleware.
• Rethought migrations: graph-based migrations that resolve branch conflicts automatically and make partial failures safe to retry, with both schema and data migrations written in TypeScript and validated against your contract.
• AI-agent friendly: your schema compiles to a machine-readable contract, every query produces a structured, inspectable plan, and middleware adds compile-time guardrails. The installers also register agent skills so your editor's AI assistant can drive Prisma Next changes safely.

How it works

Prisma Next is contract-first:

1. Author your schema: define your models in a Prisma schema (.prisma) or directly in TypeScript.
2. Emit the contract: a lightweight build step turns the schema into a deterministic JSON contract plus TypeScript types, artifacts you can read and diff.
3. Query with a composable DSL: write queries inline against the typed contract, and each one compiles to a verifiable plan at runtime with no hidden client methods.

Scaffold a new project in one command:

bunx create-prisma@next

Supported databases

Prisma Next ships first-class support for PostgreSQL (the primary target, on track for general availability) and MongoDB. SQLite is the next SQL target on deck, with MySQL to follow.

The Prisma Next blog series

Prisma Next is being built in the open. The full story spans this nine-part series:

1. The Next Evolution of Prisma ORM. The announcement that explains why Prisma is being rebuilt and tours the new query API, extensibility model, migrations, and AI-friendly design.
2. Prisma Next Roadmap. What is already built and the phased plan that runs through early access to general availability.
3. Rethinking Database Migrations. How graph-based migrations replace timestamp ordering with schema state hashes, keeping branches reconcilable and migrations explicit and verifiable.
4. TypeScript Migrations in Prisma Next. Writing schema migrations in TypeScript that are checked against your contract at compile time.
5. Data Migrations in Prisma Next. Running type-safe data transformations alongside schema changes with the same query builder you use in your app.
6. Prisma Next: A Call for Extension Authors. An invitation for the community to build databases, extensions, and integrations on the new extension API.
7. Prisma Next: April Milestone Complete, the Extension API Is Open. The extension API opens with its first community extensions (pgvector, arktype-json, ParadeDB, and CipherStash), plus progress across migrations, transactions, and MongoDB.
8. Prisma Next Early Access: Write Your Contract, Prompt Your Agent, Ship Your App. The early access launch and the agent-driven workflow it enables.
9. Prisma Next Is ~90% As Fast as Raw PG. Benchmarks showing Prisma Next reaching roughly 90% of the raw PostgreSQL driver's throughput at a fraction of the bundle size.

Next steps