Choose Between ManagedRuntime and Effect.provide

Guideline

Use ManagedRuntime.make(layer) when you run many effects with the same dependencies—HTTP servers, long-running workers, REPLs. Use Effect.provide(program, layer) for one-off scripts or when each program has different needs.

Rationale

A runtime bakes your layers into a reusable execution environment. You call runtime.runPromise(effect) without passing layers each time. For one-off scripts, providing at the edge is simpler. For servers that handle many requests with the same DB, logger, and config, a runtime avoids repeating Effect.provide on every handler.

Good Example

import { Effect, Layer, ManagedRuntime } from "effect"

class Db extends Effect.Service<Db>()("Db", {
  sync: () => ({ query: () => Effect.succeed("data") }),
}) {}

// One-off: provide at the edge
const oneOff = Effect.gen(function* () {
  const db = yield* Db
  return yield* db.query()
})
Effect.runPromise(Effect.provide(oneOff, Db.Default))

// Many runs: create runtime once
const runtime = ManagedRuntime.make(Db.Default)
const effect = Effect.gen(function* () {
  const db = yield* Db
  return yield* db.query()
})
runtime.runPromise(effect).then(console.log)
runtime.dispose()

Explanation: One-off scripts use Effect.provide at the call site. Long-running apps create a runtime and reuse it; call dispose() when shutting down.

Anti-Pattern

Creating a new runtime for every request, or using Effect.provide in a hot loop instead of building a runtime once.