OpenAI expanded Codex with cloud, voice and security tools

OpenAI announced several changes to Codex at its Dev Day event on Tuesday. It said Codex now had reusable cloud development environments that could be reached from a computer, a phone or the cloud.

The company said those environments were meant to start tasks faster and give teams a shared workspace with approved settings and permissions. It also said the Codex CLI could use voice for starting and directing tasks, and that a new /agents view helped with delegation and task tracking.

OpenAI added that Codex could be used in a code review flow inside the ChatGPT desktop app. In that flow, users could read summaries, inspect changes and ask Codex about issues before commenting on GitHub pull requests or GitLab merge requests; automatic reviews were also included.

On the security side, OpenAI introduced Codex Security Cloud for scanning GitHub repositories on demand, on a schedule or when new commits arrived. The company said Codex would investigate findings, remove duplicates and prepare fixes in the cloud, and that the toolset included models from its Daybreak Blue cybersecurity initiative without a separate application.

OpenAI also announced a Decisions API for real-time decision-making, which it said used Luna to answer user-defined questions with predefined replies. It also updated the Agents API to support computer use and Amazon Bedrock Managed Agents for building OpenAI agents that ran on AWS.