Drizzle vs Kysely — TypeScript ORM vs Query Builder: Pick Your Poison

Drizzle is a full ORM with migrations and type safety; Kysely is a lean query builder for SQL purists. One's a Swiss Army knife, the other's a scalpel.

What They Actually Do

Drizzle is a TypeScript ORM that pretends it's just a query builder—don't buy it. It handles schema definitions, migrations, and relationships while keeping your types tight. Kysely is a query builder that's basically SQL with training wheels: you write SQL-like syntax, but it's all TypeScript under the hood. If you want an ORM that doesn't feel like one, Drizzle's your pick; if you'd rather write raw SQL but with autocomplete, Kysely's the move.

Type Safety Showdown

Drizzle's type safety is runtime-enforced—your schema is your source of truth, and it catches mismatches before they hit the database. Kysely's type safety is compile-time only, meaning you can still shoot yourself in the foot with a dynamic query. Drizzle wins here because it doesn't just promise safety; it enforces it. Kysely's approach is fine if you trust yourself not to write buggy SQL, but let's be real—nobody does.

Migrations: Built-in vs DIY

Drizzle ships with a migration CLI that generates and runs migrations automatically—no more hand-rolling SQL files. Kysely makes you write migrations manually or use a third-party tool like Flyway. This is the killer feature: Drizzle saves you from migration hell, while Kysely leaves you to fend for yourself. If you've ever lost a weekend to a broken migration, you'll pick Drizzle every time.

Pricing and Ecosystem

Both are open-source and free, but the real cost is in setup time. Drizzle has a smaller ecosystem but integrates neatly with tools like Next.js and Vercel. Kysely leans on the Node.js SQL ecosystem, so you'll need to pull in extra libraries for things like connection pooling. Drizzle's all-in-one approach means less configuration; Kysely's modularity means more pieces to manage. Free doesn't mean cheap—pick based on how much time you want to spend glueing things together.

Performance and Gotchas

Drizzle uses zero-cost abstractions—it compiles down to efficient SQL, so there's minimal overhead. Kysely is lightweight by design, but you pay for that with more boilerplate. The gotcha? Drizzle's ORM features can tempt you into lazy queries if you're not careful; Kysely forces you to think about each query, which is either a pro or a con depending on your team's SQL skills. For most apps, Drizzle's performance is fine; for high-throughput systems, Kysely's control might edge it out.

When to Use Which

Use Drizzle if you're building a new app and want type safety without the ORM bloat—it's the sweet spot between convenience and control. Use Kysely if you're migrating an existing SQL-heavy app or have a team of SQL wizards who hate ORMs. Both tools are solid, but Drizzle's built-in migrations and tighter type integration make it the default choice for modern TypeScript projects.

Quick Comparison

Factor	drizzle	kysely
Type Safety	Runtime-enforced via schema	Compile-time only
Migrations	Built-in CLI	Manual or third-party
Learning Curve	Moderate (ORM concepts)	Low (SQL-like syntax)
Performance Overhead	Low (zero-cost abstractions)	Very low (lightweight)
Ecosystem Size	Small but growing	Leverages Node.js SQL libs
Boilerplate Required	Minimal (schema-driven)	High (manual queries)
Pricing	Free, open-source	Free, open-source
Best For	New TypeScript apps	SQL-heavy migrations

The Verdict

Use drizzle if: You're starting a TypeScript project and want type safety without the hassle of separate migration tools.

Use kysely if: You're working with complex, existing SQL queries or have a team that prefers writing raw SQL.

Consider: Prisma if you need a more mature ORM with a GUI and broader database support—but be ready for heavier abstraction.

The Bottom Line

Drizzle wins

Drizzle gives you a full-stack TypeScript ORM with zero-cost abstractions and built-in migrations—no more juggling separate tools. Kysely makes you write SQL by hand, which is great until you're debugging a 3AM migration mess.

Related Comparisons