Do coding agents recommend Codex code review?
Codex code review was chosen in 13% of 290 judged code review sessions, ranking third. Measured with Claude Code, Codex and Cursor.
Codex code review was chosen in 13% of 290 judged code review sessions, ranking third. It was also raised as a candidate in 4 further sessions without being chosen.
This page reports what happened when Claude Code, Codex and Cursor had to solve a problem in code review inside a realistic codebase. Not what a chat assistant says about Codex code review. What an agent actually installed.
The numbers
| Category | Code review |
| Sessions in the category | 290 |
| Sessions where Codex code review was chosen | 37 |
| Install share | 13% |
| Rank in category | 3 of 10 |
| Codebases it won in | 7 |
| Raised as a candidate, not chosen | 4 |
| Chosen when considered | 90% |
| Site | openai.com |
By agent
The three agents treat Codex code review very differently. Codex chose it in 37% of its sessions in this category and Claude Code in 0%, a spread of 37 points.
| Agent | Sessions | Chose Codex code review | Share |
|---|---|---|---|
| Claude Code | 100 | 0 | 0% |
| Codex | 100 | 37 | 37% |
| Cursor | 90 | 0 | 0% |
A single blended install share would report the midpoint and hide both ends. If most of your users run Claude Code, your real position here is 0%, not 13%.
By who was asking
Codex code review does much better with one kind of buyer than another. It won 22% of sessions asked as vibe coder and 3% of those asked as enterprise team.
| Who is asking | Sessions | Chose Codex code review | Share |
|---|---|---|---|
| Vibe coder | 58 | 13 | 22% |
| Junior developer | 116 | 10 | 9% |
| Senior engineer | 58 | 12 | 21% |
| Enterprise team | 58 | 2 | 3% |
What Codex code review was up against
The full ranking in code review, from the same sessions:
| # | Product | Runs won | Share |
|---|---|---|---|
| 1 | Claude Code review | 76 | 26% |
| 2 | Cursor Bugbot | 71 | 24% |
| 3 | Codex code review (this page) | 37 | 13% |
| 4 | CodeRabbit | 34 | 12% |
| 5 | Qodo Merge | 24 | 8% |
| 6 | Built in-house (no product adopted) | 19 | 7% |
| 7 | Semgrep | 10 | 3% |
| 8 | GitHub Copilot code review | 9 | 3% |
What this means
Agents almost never raise Codex code review without choosing it: 4 sessions raised against 37 chosen. When it gets considered, it usually wins. The constraint is how rarely it gets considered at all, which is a presence problem: templates, framework integrations and pages that answer the configuration and production questions agents actually search for.
Where these numbers come from
The 290 sessions in code review are part of a published set of 5,915, run with real coding agents inside realistic codebases and judged blind. The full method is on one page: how we measured this.
Every code review run can be replayed on the board.
If you work on Codex code review: the judge recorded a reason for every session where it was raised and passed over. Those reasons are in the transcripts.
<!-- generated by scripts/write-data-pages.mjs -->
Common questions
Do coding agents recommend Codex code review?
Yes. Codex code review was chosen in 37 of the 290 judged sessions in code review, a 13% install share, ranking third in its category.
Does Claude Code recommend Codex code review?
In 0 of the 100 sessions in code review run with Claude Code, which is 0%.
Do different coding agents treat Codex code review differently?
Yes, and by a wide margin. Codex chose it in 37% of its runs and Claude Code in 0%.
How was this measured?
Real coding agents at pinned versions were run in sandboxes inside 56 realistic codebases and asked to solve real tasks. A simulated project owner approved or questioned each recommendation before any code was written, and a judge from a model family that builds none of the agents read every session blind.
How often is Codex code review considered but not chosen?
It was raised as a candidate in 4 sessions without being chosen, and chosen in 37. That is a 90% conversion from considered to chosen.
Where this comes from
Armature ran 5,915 judged sessions with Claude Code, Codex and Cursor inside 56 realistic codebases, and published every run. The numbers on this page come from that work.