OpenAI has updated its software engineering agent, Codex, to allow developers to run reusable cloud environments from any device.
The announcement came during OpenAI’s Dev Day on Tuesday. Alongside the environment changes, the company released a refreshed Command Line Interface, a new code review feature, infrastructure hardening tools, and other API upgrades.
Previously, Codex could spin up cloud tasks remotely. This update shifts the approach by making those environments persistent and configurable rather than treating each session as an isolated sandbox.
Developers can now launch Codex on a computer, remotely via phone, or in the cloud. OpenAI states that these reusable environments speed up task initiation and provide teams with a shared workspace containing approved settings and permissions.
The Codex Command Line Interface now supports voice commands to start and direct tasks. A new /agents view simplifies delegating work and tracking multiple tasks simultaneously. Other improvements cover editing prompts, resuming sessions, and using worktrees. The terminal user interface has also been redesigned for better readability during long sessions.
Users can now access a new code review experience inside the ChatGPT desktop app. This feature lets them read summaries, explore code changes, or query potential issues before submitting feedback on GitHub pull requests or GitLab merge requests. Automatic reviews allow Codex to perform an initial check while users step away from their computers.
On security, OpenAI is rolling out Codex Security Cloud. This toolset helps developers secure their infrastructure by scanning entire GitHub repositories on demand, on a schedule, or when new commits arrive. Codex investigates findings, removes duplicates, and prepares fixes in the cloud, even when the laptop is closed.
Access to models via OpenAI’s cybersecurity initiative Daybreak Blue is included in the toolset without requiring a separate application.
Several API updates accompany these changes. A new Decisions API handles real-time decision-making by using Luna to answer user-defined questions with predefined answers. The Agents API has been updated to support computer use and the ability to build OpenAI agents that run entirely on Amazon’s Bedrock Managed Agents.
What it means
Developers can maintain a consistent development environment across different machines rather than reconfiguring settings for every session. The voice commands and improved interface reduce the friction of managing complex tasks, while the automatic security scanning ensures infrastructure health without constant manual oversight.
