OpenAI used its Dev Day keynote on Tuesday to push its software-engineering agent Codex further off the laptop, announcing reusable cloud development environments, a rebuilt command-line tool and a security product.

The headline change is that Codex's cloud environments are becoming persistent and configurable rather than one-off remote sandboxes. Developers can now run Codex from a computer, from a phone or in the cloud, and teams get a shared workspace with approved settings and permissions — which OpenAI says makes tasks start faster and keeps different machines on the same configuration.

The Codex CLI gets an update that lets developers start and direct tasks by voice, plus a new /agents view for delegating work and tracking several tasks at once. Routine interactions — editing prompts, resuming sessions, using worktrees — were reworked, and a cleaner terminal interface is aimed at long sessions.

Codex is also being pulled into a code review experience inside the ChatGPT desktop app, where users can read summaries, explore changes, or ask the agent about potential problems before commenting on a GitHub pull request or GitLab merge request. Automatic reviews can run while the developer is away from the keyboard.

On the security side, OpenAI introduced Codex Security Cloud, which scans entire GitHub repositories on demand, on a schedule, or whenever new commits land, then investigates findings, removes duplicates and prepares fixes in the cloud — even when the user's laptop is closed. The toolset includes access to models from OpenAI's Daybreak Blue cybersecurity initiative without a separate application.

The API announcements were quieter but consequential: a Decisions API that uses a model OpenAI calls Luna to answer user-defined questions with predefined answers in real time, and an updated Agents API that adds computer use and support for Amazon Bedrock Managed Agents, letting OpenAI-built agents run entirely inside AWS.

Taken together, the updates position Codex as something closer to a hosted engineering platform than an autocomplete tool.