OpenAI introduced reusable cloud development environments for its Codex software engineering agent on 29 September, TechCrunch reported. Announced at the company’s Dev Day, the update also includes voice controls, code review in the ChatGPT desktop app, security scanning and changes to two APIs.
Key points
- Codex cloud environments are becoming reusable and configurable across devices.
- Voice controls and a new desktop review experience add ways to direct and inspect Codex’s work.
- Codex Security Cloud can scan GitHub repositories when requested, on a schedule or after new commits.
- OpenAI also introduced a Decisions API and updated its Agents API.
Codex keeps cloud environments between tasks
Codex could already start remote tasks in the cloud. The change is to the environment around those tasks: instead of treating each one as a separate, temporary workspace, OpenAI is making development environments reusable and configurable. Think of leaving a workbench set up for the next job rather than clearing it after every job; the work can change while the place prepared for it remains available.
OpenAI says developers will be able to run Codex on a computer, remotely from a phone or in the cloud. The company says reusable environments are intended to help tasks start faster and to give teams a shared workspace with approved settings and permissions. Those permissions matter when work moves beyond a single remote task: a shared setup gives the next task a defined place to run, rather than requiring a separate cloud sandbox for each one.
The update therefore changes what developers building with Codex can keep ready between tasks. OpenAI describes both quicker starts and a workspace that teams can share as aims of the design, rather than results from a reported timing test.
Codex CLI adds voice and task tracking
An update to the Codex CLI will let developers start tasks and direct them by voice. It also adds an /agents view for delegating work and following several tasks at once. OpenAI has made changes to prompt editing, session resumption and worktrees, alongside a terminal interface it says is designed to make long sessions easier to read.
In the ChatGPT desktop app, a new code review experience lets users read summaries, inspect changes and ask Codex about possible issues before giving feedback on GitHub pull requests or GitLab merge requests. Automatic reviews can also provide an initial pass while users are away from their computers.
Checking a change before sharing it could involve reading a summary, looking through the altered lines and asking about a possible problem before sending feedback. The review experience puts those activities together, while the automatic option could leave an initial review ready to inspect later.
Codex Security Cloud scans GitHub repositories
Codex Security Cloud adds scans of entire GitHub repositories, which developers can request directly, schedule or trigger when new commits arrive. After a scan, Codex can investigate findings, remove duplicates and prepare fixes in the cloud, including when a laptop is closed, TechCrunch reported.
That sequence extends Codex’s remote work from carrying out development tasks to checking the resulting code for security issues. The toolset also includes access to models offered through OpenAI’s Daybreak Blue cybersecurity initiative without a separate application.
OpenAI introduced a Decisions API that uses Luna to answer questions set by users with predefined answers. It also updated the Agents API to support computer use and Amazon Bedrock Managed Agents, allowing developers to build OpenAI agents that run entirely on AWS.
