OpenAI Codex Adds Record & Replay: Turn a Demonstrated Mac Workflow Into a Reusable Skill

OpenAI added Record & Replay to the Codex app in version 26.616, a macOS feature that turns a workflow you demonstrate into a reusable skill. Instead of re-describing a multi-step task every time you want Codex to run it, you show the task once and Codex can replay it. The capability builds on Codex's Computer Use, so it inherits the same enablement and regional requirements.

This is a short, source-backed rundown of exactly what OpenAI documented, what is required to use it, and where it fits alongside the rest of Codex's desktop automation.

Key Takeaways

• What shipped: Record & Replay in Codex app 26.616 -- a macOS feature that turns a demonstrated workflow into a reusable skill, per OpenAI's Codex changelog.
• Built on Computer Use: the feature depends on Computer Use, which you or your administrator must enable.
• Platform: macOS.
• Regional availability: not available at launch in the European Economic Area, the United Kingdom, and Switzerland -- the same regional carve-out OpenAI applies to Computer Use.
• What we are not claiming: OpenAI's durable changelog does not spell out the recording commands or UI, so the specific mechanism is not detailed here.

What OpenAI Says Record & Replay Does

According to OpenAI's Codex changelog, Codex app 26.616 introduces Record & Replay, which turns a demonstrated workflow into a reusable skill. The framing is demonstration over description: you perform a workflow once and Codex captures it as something it can run again.

That is a different shape from prompting. A normal Computer Use task starts from instructions you type; a recorded skill starts from a sequence you actually carried out. For repetitive, GUI-bound work -- the kind of multi-app, click-through tasks that are awkward to specify in prose -- demonstrating once and replaying is the more natural pattern.

OpenAI announced the feature on X as well, but the changelog is the durable record of what shipped and in which version.

What You Need to Use It

Record & Replay is not a standalone toggle. Per OpenAI's changelog, it builds on Computer Use, with two requirements that follow directly:

• Enable Computer Use. You or your administrator must turn Computer Use on before Record & Replay is usable. In managed or enterprise setups, that may be an administrator action rather than something you can flip yourself.
• Run macOS. Record & Replay is a macOS feature in this release.

Availability

OpenAI states that initial availability excludes the European Economic Area, the United Kingdom, and Switzerland. This matches the regional restriction OpenAI already applies to Codex Computer Use on macOS, so the carve-out is consistent rather than new. If you are in one of those regions, the feature is not available to you at launch.

Why This Matters

Codex's desktop automation has been building toward exactly this. Computer Use gave Codex the ability to drive Mac apps; Locked computer use let those tasks keep running after your screen locks. Record & Replay adds the missing authoring step: instead of writing instructions for every GUI task, you demonstrate it once and keep it as a skill.

The most valuable Computer Use work is the "stuff your CLI can't do" category -- reproducing GUI-only issues, running multi-step flows across desktop apps, operator-style tasks that span several windows. Those are precisely the tasks that are tedious to describe and easy to show. Turning a demonstration into a reusable skill is the kind of ergonomics change that makes the repetitive cases worth automating at all.

FAQ

The questions above cover what OpenAI has documented. For anything beyond the changelog -- the exact recording commands, the editing surface, or how skills are stored -- watch OpenAI's Codex changelog for updates, since the durable page does not yet detail those mechanisms.