OpenAI has introduced a broader set of capabilities for Codex, its software engineering agent, designed to make AI-assisted development more flexible across devices and teams.
The central addition is a system of reusable cloud development environments. Instead of creating isolated remote sessions for each task, developers can now configure persistent workspaces that can be reopened from a computer, phone or cloud-based setup. These environments are intended to launch tasks faster while giving teams shared access to approved settings and permissions.
Development work beyond a single device
The refreshed Codex CLI also adds voice-based task creation and guidance. A new agents view helps users delegate work, monitor several tasks at once and manage longer programming sessions through a cleaner terminal interface. Updates to prompt editing, session recovery and worktree support aim to streamline everyday coding routines.
Within the ChatGPT desktop app, Codex can now support a new code review workflow. Developers can view change summaries, explore proposed edits and ask questions about potential issues before preparing feedback for GitHub pull requests or GitLab merge requests. Automated reviews allow Codex to conduct an initial assessment while a user is away from their workstation.
Infrastructure support and new APIs
OpenAI also unveiled Codex Security Cloud, a toolkit for reviewing infrastructure code in GitHub repositories. The service can examine repositories on demand, on a schedule or following new commits, then organize findings and prepare suggested fixes in the cloud. The offering includes models from OpenAI's Daybreak Blue cybersecurity initiative.
On the API side, OpenAI announced a Decisions API for real-time responses to user-defined questions with preset answer options. The updated Agents API adds computer-use support and enables developers to build OpenAI agents running entirely through Amazon Bedrock Managed Agents on AWS.
By shifting coding assistance toward shared, persistent and device-independent environments, these tools could help software teams build more continuous and collaborative development workflows in the years ahead.
