OpenAI Shipped a Codex Plugin for Claude Code. Read the Signal.
The official codex-plugin-cc repo lets Claude Code users run Codex reviews and delegate tasks. The interop matters more than the feature.
OpenAI now ships an official plugin that runs Codex inside Claude Code: read-only code reviews, an adversarial review mode that challenges design decisions, task delegation to background Codex jobs, and full session handoff between the two tools. Apache-2.0, version 1.0.6 shipped July 8, and it crossed 26,000 stars within days.
The feature is useful. The signal is bigger. OpenAI building for a competitor’s surface concedes that the orchestration layer, not the model, is where developers live now. For an enterprise, that changes what you standardize on: pick the orchestrator and the governance around it carefully, because the models behind it are becoming swappable.
The pattern I would actually adopt from this is cross-vendor review. Having one vendor’s agent write code and a different vendor’s agent review it reduces correlated blind spots, the same reason you do not let the author approve their own pull request. The adversarial review mode, which pressure-tests the chosen approach rather than the diff, maps directly onto the verification stage of an agentic pipeline.
One caveat before rolling it out: the optional review gate hooks every stop and can loop Claude and Codex against each other, burning usage limits fast. Leave it off. Start with background reviews on real branches and compare what each side catches.
Written by Adib Kadir. Product and engineering executive focused on rolling out AI at enterprise scale.
Start a conversation