OpenAI's Codex Now Learns Workflows by Watching You Do Them Once

OpenAI just shipped a new feature for the Codex app called Record & Replay, and it flips the usual agent-programming model on its head. Instead of writing instructions or prompts to teach Codex how to do something, you just show it once. Codex watches, learns, and turns your demonstration into a reusable, editable skill.

The feature is currently rolling out to select markets, with broader availability coming later.

Show once, reuse forever

The core idea is dead simple: you perform a task in front of Codex , filing an expense report, submitting a time-off request, navigating a multi-step internal tool , and Codex records the workflow. You control when recording starts and stops. Once you're done, Codex converts that demo into a skill: a structured, inspectable artifact that it can replay on demand.

Skills in Codex are already a first-class concept. Agent skills extend Codex with task-specific capabilities, packaging instructions, resources, and optional scripts so Codex can follow a workflow reliably. What Record & Replay adds is a new input method for creating those skills , one that requires zero writing. You demonstrate instead of describe.

What the skill looks like under the hood

The resulting skill isn't a black box. Once Codex generates it from your recording, you can inspect and edit it. Codex uses progressive disclosure to manage context efficiently: it starts with each skill's name, description, and file path, and loads the full instructions only when it decides to use a skill. This means the generated skill follows the same format as any hand-authored one , a SKILL.md file with optional scripts and references.

That's important for trust. You're not handing Codex a magic recording it can never explain. You can open the skill, see what it captured, fix anything it got wrong, and commit it to your repo so teammates can use it too. Skills stored in ~/.codex/skills are available from any repo, while skills in the current repo can be committed so teammates can use them.

The problem it actually solves

The friction point Record & Replay targets is real. Plenty of workflows are genuinely hard to describe in text , especially multi-step UI flows through internal tools, HR portals, or legacy dashboards. Writing accurate instructions for these is tedious, error-prone, and often requires you to think through every edge case upfront.

The existing alternative was to use Codex's skill-creator workflow, which involves pasting in Slack threads, docs links, or example outputs and prompting Codex to formalize them. That still works well for knowledge-heavy tasks. But for procedural, click-through workflows, a recording is a much more natural fit.

Record & Replay also connects to a broader shift in how Codex is being positioned. Codex is evolving from a developer tool into a broader AI agent for knowledge work, personal productivity, and complex multi-step tasks , reducing friction, managing information overload, and automating repetitive work. Skills are the mechanism for that generalization, and Record & Replay makes skill creation accessible to anyone who can perform a task, not just those who can write precise instructions.

Practical use-cases

The feature is most useful for tasks that are:

• Recurring and procedural , expense reports, time-off requests, weekly status updates through a portal
• UI-heavy , workflows that require clicking through multiple screens in a web app or internal tool
• Hard to describe in text , anything where the steps are obvious when you do them but awkward to write down
• Team-shared , onboarding steps, compliance workflows, or standard operating procedures that everyone on a team runs

It's less useful for tasks that are highly variable, require significant judgment, or are already well-served by a written prompt or skill. For those, the existing $skill-creator workflow remains the right tool.

Availability and access

Record & Replay is an optional feature and is currently rolling out to select markets. Broader availability is coming. It lives inside the Codex desktop app, which is available on macOS and Windows. No additional cost is mentioned beyond existing Codex plan access.

To use it, you'll need the Codex app updated to the latest version. Once the feature reaches your region, you can start a recording, walk through your workflow, stop the recording, and let Codex generate the skill. From there, the skill is available in the app's skill picker or can be invoked explicitly with $skill-name in any Codex thread.

The real test will be how well Codex generalizes from a single demo , especially for workflows with conditional branches or variable inputs. But as a way to lower the floor for skill creation, Record & Replay is a meaningful step toward making Codex useful for people who think in actions, not instructions.