> ## Documentation Index
> Fetch the complete documentation index at: https://docs.convonexai.tech/llms.txt
> Use this file to discover all available pages before exploring further.

# Deployment

> What actually runs when you publish — a real, managed, isolated service.

While you build, your app runs as a live [preview](/building/preview). When you
[publish](/shipping/publish), the deployment capability creates the production
version: a real managed service in your project's own isolated environment,
serving your [production URL](/capabilities/domains-and-urls).

## What publishing provisions

* A production instance of your app, separate from the preview — the preview can
  sleep, restart, or be mid-change without affecting the published app.
* Production configuration: the same database, storage, and secrets your app was
  built and verified against, injected the same way.
* Health checks, so the platform notices if the app stops responding.

## Operating the published app

Ask the agent — it can:

* **Redeploy** after changes (publish again from the workspace),
* **Restart** the production app with a rolling restart (zero downtime),
* **Scale** the number of running instances up or down,
* **Read production logs** when you're chasing a live issue,
* **Check status** — which revision is running and whether it's healthy.

## Good to know

* The production app runs with production settings (`NODE_ENV=production`),
  resource limits, and health probes — it is not the dev preview with a
  different URL.
* Publishing again rolls the new version out in place. There is no self-service
  one-click rollback yet — if a publish goes wrong, the fastest path is asking
  the agent to fix forward or re-publish the previous state, and support can
  help in an emergency.
* The published app holds no platform credentials and no cloud credentials; it
  reaches capabilities through the platform's scoped runtime broker.