Codex Desktop: Codex Helper (Renderer) high CPU when popout window shortcut is enabled

Resolved 💬 2 comments Opened Apr 28, 2026 by mothieras Closed Apr 29, 2026

This is a Codex Desktop macOS app issue, not a Codex CLI issue.

What version of the Codex App are you using?

Codex Desktop 26.422.30944 (CFBundleVersion 2080)

What platform is your computer?

macOS 26.4.1 (25E253), arm64

What issue are you seeing?

Enabling the popout window shortcut appears to be strongly correlated with sustained high CPU usage in Codex Helper (Renderer) while Codex Desktop is otherwise idle.

In local testing, disabling the popout window shortcut returns CPU usage to normal. When the shortcut is enabled, Codex Helper (Renderer) can consume more than two CPU cores continuously.

The exact trigger timing seems inconsistent:

  • In one test, enabling the shortcut did not immediately cause abnormal CPU usage until Codex Desktop was quit and relaunched.
  • In another test, it looked like CPU usage became abnormal shortly after enabling the shortcut.

So this may be related to the shortcut feature initialization or renderer-side event handling, but I cannot yet say whether app restart is always required.

What steps can reproduce the bug?

  1. Launch Codex Desktop on macOS.
  2. Enable the popout window shortcut.
  3. Observe Codex Helper (Renderer) CPU usage in Activity Monitor.
  4. If CPU usage remains normal, quit and relaunch Codex Desktop.
  5. Observe Codex Helper (Renderer) CPU usage again.
  6. Disable the popout window shortcut.
  7. Observe CPU usage returning to normal.

What is the expected behavior?

Enabling the popout window shortcut should not cause sustained renderer CPU usage while Codex Desktop is idle.

Additional information

This looks distinct from active model execution because the issue appears while idle and changes with the popout shortcut setting. It may be related to the global shortcut listener, popout window pre-rendering, or renderer-side event loop after loading the persisted shortcut setting.

View original on GitHub ↗

This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗