Deployment

Two ways to run a scaffolded project in production — pick Docker for a container platform, PM2 for a plain VM you manage directly.

Docker

docker compose up --build

Builds the HTTP adapter into a container and starts it alongside Redis. For production, push the image to any container platform with REDIS_URL and your model/tool API keys set as environment variables.

PM2

For a bare VM without a container platform, run the HTTP adapter under PM2 instead — it handles restart-on-crash, log capture, and boot persistence:

npm install -g pm2
pm2 start "npx loopengine serve" --name loopengine
pm2 save
pm2 startup   # wires PM2 into this machine's own boot sequence

loopengine serve (not dev) is the right command here — it's the same server with no tsx --watch file-watching, which would otherwise fight with PM2's own restart supervision. serve also loads .env automatically, equivalent to --env-file-if-exists=.env, so ANTHROPIC_API_KEY and friends just need to be in the project's .env file.

For a checked-in config instead of a one-off command, an ecosystem.config.cjs in the project root:

module.exports = {
  apps: [
    {
      name: 'loopengine',
      script: 'npx',
      args: 'loopengine serve',
      env: { NODE_ENV: 'production' },
    },
  ],
}

pm2 start ecosystem.config.cjs
pm2 logs loopengine

Set REDIS_URL in .env too once you're running more than one PM2 instance (e.g. PM2's own cluster mode) against the same sessions — see Basic Configuration for the full environment variable list.