Codex changelog: where it lives and what to change
Where the official Codex changelog lives, how to read stable and alpha releases, and a routine for what to review in your setup after each Codex update.
The Codex changelog is split across a few official places, and none of them tells you what to change in your own setup. This guide shows where the Codex changelog lives, how to read a Codex release, and a short routine to run after each update. Checked on 2026-10-10.
Where the Codex changelog lives (and which source to trust)
- Codex changelog on OpenAI Developers: the official product changelog for Codex. The page is rendered with JavaScript, so open it in a browser rather than scraping it.
- openai/codex releases on GitHub: per-version notes for the CLI.
- CHANGELOG.md in the repository: it states that the changelog can be found on the releases page, so it is a pointer, not a second source. Its “Unreleased” section is not a list of shipped changes.
Third-party mirrors of the changelog exist. Treat them as copies that may lag, and link to the official pages instead.
How to read a Codex release: stable vs alpha tags
On the releases page, tags look like rust-v0.162.1 (stable) and rust-v0.163.0-alpha.N (pre-release). On 2026-10-10 the latest stable tag we saw was rust-v0.162.1, published 2026-10-09; newer versions will have shipped by the time you read this.
- Read stable releases for what you will actually get.
- Treat alpha tags as previews: notes can change, and the build is not meant for pinned environments.
- Scan each entry for anything that touches how you run Codex: configuration, approvals and sandboxing, authentication, and command-line flags. Skip the rest.
Check which version you run and update safely
Run codex --version to see your installed version (confirm the flag with codex --help on your build). The README lists the supported install methods: the install script, npm install -g @openai/codex, brew install --cask codex, or a binary from the latest GitHub release. Update with the same method you installed with, so you do not end up with two copies on your PATH. The README also says the standalone installers fall back to GitHub Releases when the primary download is unavailable.
A repeatable routine for each update
- Note your current version.
- Read every stable entry between it and the target version on the changelog or the releases page.
- Mark entries that touch config, approvals, sandboxing, auth or flags.
- Update on one machine, or one CI branch, first.
- Run a small known task and compare behavior before and after.
- Roll out, and record the version you pinned.
What to change in your setup
- CI: pin a specific stable version instead of “latest”, and bump it deliberately after reading the notes.
- Configuration: re-check your Codex config file against any release entry that mentions config changes.
- Instruction files: if you keep an AGENTS.md, re-read it after updates that mention agent behavior, and remove workarounds the release says are fixed. For the same idea on the Claude side, see Claude Code hooks examples and Claude Code skills.
- Team docs: write down the pinned version and the install method.
- Do not guess: if a release note is unclear, test it on a branch instead of assuming a behavior.
FAQ
- Where is the official Codex changelog?
- On the OpenAI Developers site at developers.openai.com/codex/changelog. Per-version notes for the CLI are on the openai/codex GitHub releases page, and the repository's CHANGELOG.md points there.
- Are alpha tags safe to install in CI?
- GitHub marks them as pre-releases. Pin CI to a stable tag and try alpha builds only on a throwaway branch.
- How do I update the Codex CLI?
- Re-run the method you installed with. The README lists the install script, npm (npm install -g @openai/codex) and Homebrew (brew install --cask codex). Check the release notes first.