one serve

Serve your web apps for production with one serve, after you’ve successfully run one build.

It takes the following arguments:

--host          # string, set the hostname to bind to
--port          # string, set the port to bind to
--compress      # boolean, enable gzip compression, defaults to true
--loadEnv       # boolean, whether to load .env files before running
--cluster       # enable cluster mode using all CPU cores
--cluster=N     # enable cluster mode with N workers

You can also set the port with the ONE_PORT or ONE_FORCE_PORT environment variables. See Environment Variables for details.

One comes with a serve command that is powered by Hono.

If you are only using SPA and SSG routes and aren’t using loaders, you don’t need to use one serve if you don’t want to - you can statically serve the results of dist/client using any server you’d like.

You must run a one build before serve.

Cluster Mode

For production deployments under heavy load, one serve supports cluster mode. This forks multiple Node.js worker processes that share the same port, distributing requests across CPU cores.

one serve --cluster       # use all CPU cores
one serve --cluster=4     # use 4 workers
one serve                 # single process (default)

Each worker handles requests independently. If a worker crashes, it automatically restarts with backoff protection to prevent crash loops.

• High-traffic production deployments
• Servers with many CPU cores
• Handling 200+ concurrent connections
• SSR-heavy workloads where renders block the event loop

• Development (use one dev instead)
• Low-traffic sites where single process is simpler and has less overhead
• Less than ~100 concurrent connections, where single process may be faster due to less IPC overhead

At 500 concurrent connections (production-style load test):

• Single process: ~2,150 req/s
• 8 workers: ~2,939 req/s (1.4x improvement)
• At 200 connections, cluster is 2.4x faster than single process

Programmatic Usage

You can also run your production server programmatically:

import { serve } from 'one/serve'

await serve()

Which takes options:

type ServeOptions = {
  // you can pass in your own Hono server
  app?: Hono

  host?: string
  port?: number
  compress?: boolean

  /**
    * Whether to run the Vite logic to load .env files before running the server
    * @default false
    */
  loadEnv?: boolean

  /**
    * Enable cluster mode with multiple worker processes.
    * Pass true to use all CPU cores, or a number for specific worker count.
    */
  cluster?: boolean | number
}