codex remote-control starts stale standalone app-server 0.132.0 while active CLI is npm-managed 0.133.0

Environment

• OS: Ubuntu 22.04.5 LTS
• Active shell CLI after npm upgrade:

``text
codex-cli 0.133.0
``

• Active shell Codex path:

``text
/root/.local/lib/node_modules/@openai/codex/bin/codex.js
``

Observed behavior 1

Running:

codex remote-control

printed:

Starting app-server with remote control enabled...
Error: Remote control is enabled on host but the connection is errored.

Workaround

Running:

codex remote-control stop
codex remote-control start

fixed the immediate connection error:

Stopping remote control...
Remote control stopped.

Starting app-server daemon with remote control enabled...
This machine is available for remote control on the host.

However, the daemon reported:

Daemon used app-server:
  path: /root/.codex/packages/standalone/current/codex
  version: 0.132.0

Expected behavior

After upgrading the active CLI to 0.133.0 via npm, codex remote-control start should either:

1. start the app-server using the same Codex executable/version as the invoking CLI/PATH
2. update the standalone app-server to the same version
3. clearly explain that remote-control intentionally uses a separate standalone app-server and how to update it

Actual behavior

There is a version split:

Active shell CLI:
  /root/.local/lib/node_modules/@openai/codex/bin/codex.js
  codex-cli 0.133.0

Remote-control app-server daemon:
  /root/.codex/packages/standalone/current/codex
  codex-cli 0.132.0

Why this matters

The user believes they are running Codex 0.133.0, but remote-control is actually served by 0.132.0. This can hide fixed bugs, introduce inconsistent behavior, and make debugging remote-control failures difficult.

Questions

• Is codex remote-control expected to follow the invoking executable / PATH?
• Or is remote-control expected to manage a separate standalone app-server under ~/.codex/packages/standalone?
• If the latter, what is the supported update path for that app-server?
• Should codex remote-control start warn when the app-server version differs from the active CLI version?

Suggested fix

• Make remote-control daemon resolution follow the invoking Codex executable when possible.
• Or explicitly document and print that it uses a separate standalone app-server.
• Add a version mismatch warning, for example:

``text
Warning: active CLI is 0.133.0 but remote-control app-server is 0.132.0.
``

• Add a command or documented flow to update the remote-control app-server.
• Improve recovery when codex remote-control reports:

``text
Remote control is enabled on the host but the connection is errored.
``

The successful workaround was:

``bash
codex remote-control stop
codex remote-control start
``