mrkeyoor.com_
Sat 12 Sept 06:29 UTC
Dataevaluationupdated 12 Sept 2026

orm review

Prisma ORM is a TypeScript toolkit for defining how an application talks to its database and for generating the code that performs those reads and writes. The Prisma 8 release candidate focuses on PostgreSQL and MongoDB, with project scaffolding, reusable agent instructions, and an extension system intended to reduce custom database plumbing.

trackingstars / 7d
Verdict

Our build completed in 130 seconds, but the test command reached the 900-second cap with 108 passed, 0 failed, and 16 skipped. Prisma 8 looks unusually deliberate about agent-guided database work and extension authorship, yet it is still a release candidate with a narrower database list than many established ORMs. Start new PostgreSQL or MongoDB experiments with it if Node.js 24 and release-candidate change are acceptable; keep production migrations gated behind your own integration tests until 8.0.0 is final.

We ran it

Install✓ · 52s809 packages · 1412 MB
Build✓ · 130s
Tests✗ timed out · 900s108 passed · 0 failed · 16 skipped of 124 (vitest)
Repo7329 files~841,419 lines of source · 68.4 MB · 17 CI workflows · tests dir

Answers from our run

Does orm build from source?

Dependencies installed in 52 seconds (809 packages), and the build succeeded in 130 seconds. We cloned commit e88e248 into a clean Debian container with 3 CPUs and no project-specific setup.

Do orm's tests pass?

Yes: 108 of 124 passed when we ran the project's own test command (vitest). Some failures need services or credentials a bare container does not have.

Who should not use orm?

Teams that require a final, stability-locked Prisma 8 release today

What are the alternatives to orm?

Drizzle ORM, Sequelize, TypeORM. Our build completed in 130 seconds, but the test command reached the 900-second cap with 108 passed, 0 failed, and 16 skipped.

Setup3/5Install worked, but 809 packages and 1412 MB are substantial
Docs4/5Clear paths for new apps, migrations, skills, and extensions
Community4/547,601 stars and a recent push, tempered by 2,610 open issues
Maturity3/5Prisma 7 is supported, while Prisma 8 remains an RC

Who it’s for

TypeScript teams starting a new PostgreSQL or MongoDB application
Teams that want database setup and query workflows documented for coding agents
Extension authors building database, validation, search, or security integrations
Existing Prisma 7 users willing to test an incremental migration

Who it’s NOT for

Teams that require a final, stability-locked Prisma 8 release today
MySQL users, because support follows SQLite and is not available in the stated support list
SQLite production projects, because support is only a proof of concept today
Small projects that cannot absorb a large dependency install and monorepo-style tooling
Teams pinned below Node.js 24, which the Prisma 8 README lists as the minimum

Setup reality

Our run installed successfully in 52 seconds, adding 809 packages and using 1412 MB on disk, then built successfully in 130 seconds. The test command did not finish before the 900-second cap: Vitest reported 108 passed, 0 failed, and 16 skipped out of 124 when time expired. That is heavier than the README's short npm create prisma or npx prisma orm init path suggests, although our lab cloned the full 7,329-file monorepo rather than scaffolding a normal application, and it ran in a Node 22 image while the README requires Node.js 24 or newer.

Prisma 8 changes the shape of a familiar ORM

Prisma ORM is a major TypeScript database project, with 47,601 GitHub stars and roots going back to 2019. This repository presents Prisma 8 as a TypeScript rewrite built around composability, extensions, and instructions that coding agents can follow. The underlying job remains familiar: give applications a consistent way to describe data and make queries without scattering database plumbing across the codebase. Prisma now wants the database contract, generated runtime, command-line workflow, and agent guidance to travel together.

The timing matters because Prisma 8 is a release candidate, while Prisma 7 remains supported on a separate branch. The README expects 8.0.0 final in four to eight weeks, warns that candidate releases may contain breaking changes, and promises upgrade recipes. It also says Prisma 7 will receive bug fixes for twelve months after the final release. That makes evaluation reasonable, but conservative production teams should keep the final release and their own regression results as explicit adoption gates.

Setup is clear, but database support is narrow

For a new app, npm create prisma offers 8 framework templates, including Next.js, Hono, Nuxt, Astro, NestJS, SvelteKit, TanStack Start, and Elysia. It can produce a runnable PostgreSQL or MongoDB app with a starter contract and agent skills. Existing projects use npx prisma orm init, which writes configuration, creates contract and database files under src/prisma/, installs the runtime, and emits the contract. A separate skills command synchronizes the agent instructions. These are specific, task-oriented entry points.

There are firm limits behind that polished path. Node.js 24 or newer is required. PostgreSQL is the primary target and MongoDB is first-class, but SQLite is only a proof of concept, and MySQL comes later. Teams standardized on either missing database should choose something else today. The repository also writes agent guidance into 4 runtime-specific locations. That may make automated work more repeatable, but generated schema and migration changes still need normal code review, staging, backups, and database-specific checks.

What happened when we ran it

We cloned commit e88e248 into our unprivileged Debian sandbox with 3 CPUs, 8 GB of RAM, no secrets, and a Node 22 image. The repository held 7,329 files, roughly 841,419 source lines, and occupied 68.4 MB before dependencies. Installation succeeded in 52 seconds through pnpm, bringing in 809 packages and expanding disk use to 1,412 MB. The build also succeeded in 130 seconds. These are full-monorepo results, not a benchmark of a generated application, but they show that the checkout installed and compiled on our box.

The complete test command did not finish within our 900-second cap. At cutoff, Vitest reported 108 passed, 0 failed, and 16 skipped out of 124. The final log showed database-state and graph-divergence integration scenarios continuing, including successful cases taking 8,457 ms, 10,488 ms, and 9,201 ms. The evidence does not show an assertion failure, so the finding is a timeout with unfinished work, not a failed assertion. Contributors should budget more than the 15 minutes our run allowed for full verification.

Extensions and version-matched agent guidance stand out

Prisma 8 makes extension authors first-class participants. The README says even PostgreSQL support uses the public SPI offered to outside authors. Listed packages cover pgvector similarity search, PostGIS geometry, experimental ParadeDB full-text search, experimental Supabase tables and role-bound clients, arktype validation for JSON columns, and CipherStash searchable encryption. That range gives the extensibility claim substance. It points toward a small core with capabilities added at defined seams, instead of forcing every uncommon database feature into the central ORM.

Agent support is equally specific. Installers leave a prisma-8.md primer at the project root and version-matched SKILL.md files for Claude Code, Cursor, Copilot Agent, Devin, and other runtimes. The README example asks an agent to add a posts model, relate it to users, and load each user's 3 most recent posts. The useful point is not replacing review. It is reducing the chance that an assistant follows instructions written for the wrong major version.

Active development does not erase release-candidate friction

The contributor footprint we measured was 1,412 MB, and the complete tests outlasted 900 seconds. The README is candid about candidate breakage and links a feature scoreboard, yet users must also reconcile a visible latest tag of v0.17.0 from August 4 with branding centered on Prisma 8. The supplied evidence does not explain that relationship, so verify package versions before adoption. Project activity is current, with a push on September 11 and a supplied release about 5 weeks old, but 2,610 open issues remain a substantial triage burden. Stars show reach, not response quality.

It belongs in the application data layer, behind safeguards

In a real stack, Prisma 8 sits between TypeScript services and PostgreSQL or MongoDB. Its generated contract and runtime belong beside application code, while credentials stay in the deployment environment and schema changes pass through review and staging. The extension SPI is most useful for needs such as vector, geometry, full-text, validation, or encryption. For a new PostgreSQL service on Node.js 24, the scaffolder and version-matched guidance justify a serious trial. For an existing Prisma 7 app, migrate incrementally. For MySQL or production SQLite, choose an alternative now, and promote the RC only after your own tests finish cleanly.

Alternatives

ProjectWhat it isPick it when
Drizzle ORMA TypeScript ORM with a SQL-like query style and schema definitions in code.pick this instead when your team wants queries that stay visibly close to SQL.
Sequelize gh↗A long-running Node.js ORM for relational databases.pick this instead when broad SQL-database support matters more than Prisma 8's agent workflow.
TypeORM gh↗A TypeScript and JavaScript ORM built around entities, repositories, and decorators.pick this instead when an entity-oriented data model fits your existing application architecture.
MikroORMA TypeScript ORM centered on entities and unit-of-work patterns.pick this instead when identity maps and unit-of-work behavior are core design requirements.

What people are saying

  1. [velocity-scout] prisma/orm

Sources

  1. Prisma ORM GitHub repository
  2. Prisma homepage

More data reviews

Lean · bokeh · datahub · open-source-games · lx-music-source · postgrest · the whole board →