OpenAI Expands Codex with Cloud Environments and Security Tools

OpenAI rolls out a wave of new features around Codex. On the agenda: reusable cloud development environments, an automated security suite, a real-time decision API, and a revamped CLI. The company presented these developments during its Dev Day, as part of a broader set of enhancements for the software engineering agent.

Dedicated Decision API and Execution on AWS

OpenAI introduces a Decisions API dedicated to real-time decision-making. This interface relies on Luna to respond to a set of user-defined questions using predefined answers. The company also updates its Agents API, which now supports the use of computers. This version includes the ability to leverage Amazon Bedrock Managed Agents to build OpenAI agents that operate entirely on AWS. These announcements are part of a larger set of API and infrastructure improvements.

Automated Security with Codex Security Cloud

OpenAI launches Codex Security Cloud, a new suite of tools designed to help developers secure their infrastructure. Teams can initiate scans of entire GitHub repositories on demand, according to a regular schedule or upon the arrival of new commits. Codex then examines the results, eliminates duplicates, and prepares fixes in the cloud, even when the user is inactive and their laptop is closed. The offering includes access to models from the Daybreak Blue cybersecurity initiative, without a separate application process.

Persistent and Shared Cloud Environments

Until now, Codex could launch tasks remotely, but OpenAI now makes its cloud environments persistent and configurable beyond simple ephemeral sandboxes. Developers can run Codex on a computer, from a phone, or directly in the cloud, with reusable environments accessible from any device. According to OpenAI, these spaces are designed to accelerate task initiation and provide teams with a shared workspace with approved settings and permissions. These capabilities arrive as part of a coherent set of improvements for Codex, presented during OpenAI's Dev Day on a Tuesday.

A More Directive CLI and Integrated Code Review

The updated Codex CLI allows users to launch and control tasks by voice. A new /agents view facilitates the delegation and simultaneous tracking of multiple jobs, while improvements target prompt editing, session resumption, and the use of worktrees. OpenAI is also beginning to roll out a more readable terminal interface for long sessions. In terms of software quality, Codex integrates into a review experience within the ChatGPT desktop application: users can view summaries, explore code changes, or query the agent about potential issues before commenting on GitHub pull requests or GitLab merge requests. Additionally, automatic reviews enable Codex to perform an initial pass when the user steps away from their workstation.