Codex CLI installs a persistent, self-updating background daemon by default

Your command-line tool now persistently runs and self-updates in the background without your explicit consent.

The Codex CLI, version 0.157.0, now installs a persistent, self-updating background daemon by default. Upon launching an interactive `codex` session, it automatically starts this daemon, which runs independently of the CLI window and persists even after all Codex windows are closed. The daemon utilizes a separate copy of the platform package, approximately 374 MB, which is distinct from the version initially installed via package managers like npm. A detached updater process also runs in the background, checking for new releases hourly and restarting the daemon, potentially disrupting active sessions. This behavior is not opt-in and users must manually disable it or stop the daemon and its updater. The daemon also retains environment variables from the first session it was launched in, which can affect subsequent sessions.

This change shifts the default user experience for the Codex CLI from a self-contained application to one with a continuously running, independently updated backend service. The automatic daemonization and self-updating mechanism bypasses user-level package management, creating a shadow installation that complicates predictable execution and resource management. Users must now actively opt-out of this background process, fundamentally altering the perceived control over their local development tools.

The automatic background daemon and self-updating behavior introduces a new layer of complexity and potential instability for users of the Codex CLI. By decoupling the backend from the installed CLI version and managing its updates independently, OpenAI introduces a point of failure and a lack of transparency for users accustomed to predictable package management. The persistent nature of the daemon and its updater consumes resources and can unexpectedly interfere with active sessions or environment configurations, necessitating user intervention to maintain control.