Command

The Command provider puts local child processes into the Stack’s dependency graph: Build for a command that produces an output asset, Exec for a command run purely for its side effects, and Dev for a long-lived dev server. There is no setup — no credentials, nothing to configure. Command.providers() is already merged into both AWS.providers() and Cloudflare.providers(), so big-hub users have it for free; a standalone Stack registers it with { providers: Command.providers(), state: Alchemy.localState() }. Import from alchemy/Command.

Resources

Build runs a command that must produce an output asset (outdir) and tracks it in state:

const build = yield* Command.Build("vite-build", {
  command: "npm run build",
  cwd: "./frontend",
  outdir: "dist",
});
yield* Console.log(build.outdir); // path to the dist directory, relative to process.cwd()
yield* Console.log(build.hash.output); // hash of the output files (when memo is enabled)

Inputs are content-hashed by default, so an unchanged project skips the rebuild entirely — see Memoization for how the hash is computed and how to scope it.

Exec runs a command purely for its side effects:

yield* Command.Exec("codegen", {
  command: "npm run codegen",
  cwd: "./packages/api",
});

yield* Command.Exec("migrate", {
  command: "npm run db:migrate",
  env: {
    DATABASE_URL: Redacted.make("postgres://..."),
  },
});

There is no output contract: the resource succeeds iff the command exits with code 0, is skipped when its inputs are unchanged, and re-runs when the command, env, or memoized files change — e.g. a recreated database’s connection URL with identical files. Deleting an Exec never reverses its side effects.

Dev runs a long-lived process scoped to the stack instance:

const dev = yield* Command.Dev("Frontend", {
  command: "npm run dev",
});
yield* Console.log(dev.url); // e.g. "http://localhost:5173"

Dev only runs during alchemy dev — during alchemy deploy it is a no-op. See Dev servers for the process lifecycle, restarts, and URL extraction.

Caution

Command strings are whitespace-split into a binary and arguments — there is no quoting support unless you set shell. Build’s delete removes the output directory (Exec’s delete does nothing). Redacted env values are unwrapped at spawn time and kept out of logs and state.

Pre-beta.58 Build.Command / Build.DevServer state converges automatically to Command.Build / Command.Dev on the next deploy (beta.58).

Compose with your cloud

The static-site guides for Cloudflare and AWS create a Command.Build (and, in dev mode, a Command.Dev) under the hood, and alchemy dev is the loop that hosts Dev processes.

Reference

Build · Dev · Exec