// the take
Strawberry is a Python GraphQL server library that uses dataclasses and type annotations as the schema definition. It's the cleanest way to write a GraphQL API in Python today — your Python types are your schema, no separate SDL file needed. Aimed at teams already using FastAPI, Django, or Starlette who want GraphQL without a separate schema language.
- Type annotations are the schema — you write normal Python dataclasses and Strawberry derives the GraphQL types, which means mypy can actually check your API contract end-to-end
- First-class async support throughout, including subscriptions over WebSocket, which puts it ahead of Graphene for async-heavy FastAPI or Starlette apps
- Built-in DataLoader support with a clean async API handles the N+1 problem without bolting on a separate package
- Apollo Federation support is real and actively tested via a dedicated federation-compatibility CI workflow, not an afterthought
- Still marked 'experimental' for Pydantic integration, which is the first thing most FastAPI users will reach for — the two-type dance (Pydantic model + Strawberry type) you're forced into is friction that Graphene avoided by just not caring about validation
- Breaking changes are frequent enough that there's a whole docs section just for them, with 16+ entries going back to 0.146.0 — upgrading a production schema requires reading changelogs carefully
- The codegen story (query-codegen to TypeScript/Python) is underdeveloped compared to the GraphQL Code Generator ecosystem; fine for greenfield but not a replacement for teams already invested in codegen tooling
- N+1 protection is opt-in via DataLoaders — there's no automatic batching or query analysis that warns you when you've missed one, so you can ship slow resolvers without the framework telling you
We read the README, the directory tree and a key source file in June 2026. Nobody pays to be listed. How we judge →