Feature-Sliced Design in monorepos: frontends that scale

How FSD disciplines layer boundaries and why it shines when web and mobile share the same domain.

Frontend monorepos tend to become a tangled web of imports as they grow. You start by sharing a button component, then a utility function, and soon domain layers are coupled to UI layers. Feature-Sliced Design (FSD) exists to prevent exactly that.

The core idea is to split code into layers with unidirectional dependencies: shared → entities → features → widgets → pages → app. An upper layer can import from a lower one, but never the reverse. It sounds restrictive, but that restriction is what makes refactors safe.

On one project, we used FSD inside a Yarn Workspaces monorepo with React web and React Native. Domain entities — user, work shift, report — lived in shared packages, while features such as "clock in" or "generate report" had their own slices on each platform. The result was that data model changes propagated predictably, with no surprises in the UI.

The biggest practical gain was onboarding. New developers quickly understand where things live and, more importantly, where they should not put code. The architecture becomes self-documenting.

FSD is not a silver bullet. For very small prototypes it can feel bureaucratic. But when you have multiple teams, multiple platforms or expected growth, it pays back the initial investment with interest.


Working on something similar? Tell me about your project.