Correction, October 11, 2026: This release uses --dangerously-bypass-approvals-and-sandbox to skip approvals and sandbox restrictions. The automatic-review option retains both a review step and restricted project access.
What happened
OpenAI shipped Codex version 0.147.0 on August 7. Codex is OpenAI’s command-line coding agent, the closest tool it has to Claude Code: you run it in a terminal and it reads, writes, and executes code on your machine.
The headline change is a new flag: --approve-for-me. Codex asks for approval when an action requires it under the configured policy and sandbox restrictions. The new flag changes that. According to the public code change on GitHub that added it, --approve-for-me routes approval requests “through automatic review” instead of asking you directly.
This differs from --dangerously-bypass-approvals-and-sandbox. That option skips confirmation and sandboxing, the restrictions that limit which files and system resources commands can reach. The release’s command-line settings retain restricted project access with --approve-for-me, while handing approval review to an automated process. Inspect the approval and action history to see what ran automatically. Keep manual approvals for sensitive tasks.
The same release tightens control in a different spot. Codex now requires you to explicitly confirm you trust a project folder before it will act on it, if it doesn’t recognize that folder already. It also improves redaction of secrets and bearer tokens, credentials that let an app act on your behalf without you typing a password each time, to reduce their exposure in terminal output and command logs.
Why it matters
--approve-for-me exists to remove a real annoyance: babysitting a stream of approval prompts during a long coding session. That’s a legitimate problem, and an automated review step is a reasonable middle ground between approving every command by hand and turning approvals off entirely.
Automatic review still needs scrutiny. Claude Code shipped two rounds of permission-check patches this week after finding that its own approval prompts could be evaded or made to show less than what would actually run. The lesson from that isn’t that Claude Code is uniquely risky. It’s that an approval step is only as good as your ability to trust what it’s actually checking. Inspect which actions run automatically rather than assuming the reviewer catches every risk.
Who should care
Anyone running Codex who finds themselves approving the same categories of command over and over, especially in unattended workflows that run without anyone watching each step. If repeated prompts are interrupting your work, --approve-for-me retains a review step and sandbox restrictions. Bypassing approvals and sandboxing removes those controls; automatic review still does not guarantee every action is safe.
What builders should do next
Keep manual approvals for work involving real credentials, production access, or anything you cannot afford to lose. Test --approve-for-me in a disposable project first, then inspect its approval and action history.
Ask Codex to add a feature and its automated tests in two separate disposable copies of the same project. Use --approve-for-me in one and the default manual approval in the other. Compare three things: how many commands the automatic review let through that you would have blocked by hand, how much time the automated version saved, and whether both runs produced code that passed the same tests. If automatic review approves something you would have stopped, investigate that action before relying on the setting for similar work.
To update, run npm install -g @openai/codex@latest if you installed Codex through npm, brew upgrade --cask codex if you installed it through Homebrew, or codex update to use Codex’s own built-in updater. Then confirm with codex --version that you’re on 0.147.0 or later.
End of article