At its Dev Day, OpenAI unveiled multiple upgrades to its software engineering agent Codex aimed at making it more useful beyond a developer’s local machine. The changes include reusable cloud development environments, a refreshed CLI, a new code review experience, infrastructure-hardening tools, and several API enhancements.
Reusable, persistent cloud environments
Previously Codex could launch cloud tasks to run work remotely; the new announcements make those cloud environments more persistent and configurable rather than treating each task as an isolated remote sandbox. Developers will be able to run Codex where they need it — on a computer, remotely from a phone, or in the cloud.
OpenAI says the reusable development environments are intended to speed task startup and provide teams with a shared workspace that enforces approved settings and permissions.
Updated Codex CLI and UI improvements
The Codex CLI has been updated to let developers start and direct tasks using voice commands. A new /agents view is designed to simplify delegating work and tracking multiple concurrent tasks. Other quality-of-life changes include improved prompt editing, session resumption, easier use of worktrees, and a cleaner terminal user interface to improve readability during longer sessions.
Code review integration in the ChatGPT desktop app
Codex is being integrated into a new code review experience inside the ChatGPT desktop application. Users can read summaries, inspect code changes, or ask Codex about potential issues before leaving feedback on GitHub pull requests or GitLab merge requests. Automatic code reviews also allow Codex to perform an initial pass while users step away from their computers.
Codex Security Cloud for repository scanning and fixes
On the security front, OpenAI introduced Codex Security Cloud, a set of tools meant to help developers secure their infrastructure. The tools can scan entire GitHub repositories on demand, on a schedule, or when new commits arrive. Codex will investigate findings, deduplicate results, and prepare fixes in the cloud — even when the user is inactive and their laptop is closed.
Notably, the toolset includes access to models from OpenAI’s cybersecurity initiative Daybreak Blue without requiring developers to submit a separate application.
API updates: Decisions API and expanded Agents API support
OpenAI also announced several API updates. A new Decisions API supports real-time decision-making: it uses the Luna model to answer a set of user-defined questions with predefined answers. The updated Agents API now supports computer use and the ability to use Amazon’s Bedrock Managed Agents to build OpenAI agents that run entirely on AWS.
Summary
The Dev Day announcements make Codex more flexible, shareable, and security-aware through persistent cloud workspaces, voice-enabled and UI improvements to the CLI, integrated code review in the ChatGPT desktop app, the Codex Security Cloud for scanning and preparing fixes, and API additions such as the Decisions API and expanded Agents API support.
