OpenAI Tuesday gave its Codex coding agent reusable cloud development environments that follow developers across devices, a voice-controlled command-line tool and a security service that scans code on a schedule, TechCrunch reported. The updates were announced at the company's DevDay event alongside new agent and decision-making APIs.

What Happened

The centerpiece is a change to how Codex runs in the cloud. Instead of spinning up a one-off sandbox for each task, developers can now configure environments that persist and can be reached from a computer, a phone or a cloud platform. Teams can bake approved settings and permissions into those environments, which OpenAI says lets tasks start faster and keeps collaborators on the same configuration.

The Codex CLI was refreshed as well. Developers can now start and steer tasks by voice, and a new /agents view lets them hand off work and monitor several tasks running in parallel. OpenAI also cleaned up the terminal interface for long sessions.

RelatedClaude Code Desktop Preview Flaw Could Have Exposed Developers’ Cloud Keys, Researchers Say

Code review moves into the ChatGPT desktop app. Users can read summaries, step through changes and ask Codex about possible problems before sending feedback to GitHub or GitLab. Automatic reviews let Codex take a first pass on GitHub pull requests and GitLab merge requests without being prompted, including while the developer is away.

On the security side, Codex Security Cloud scans GitHub repositories on demand or at set intervals. It investigates what it finds, removes duplicate findings and prepares fixes in the cloud on its own. The service includes access to models from Daybreak Blue, OpenAI's cybersecurity effort, without the separate application process those models otherwise require.

OpenAI also announced a Decisions API for real-time decision-making that uses its Luna model, an Agents API update that adds computer use, and integration with Amazon Bedrock Managed Agents so customers can build OpenAI-powered agents entirely on AWS infrastructure. TechCrunch's report did not include pricing or availability dates.

Why It Matters

Persistent environments address one of the practical limits of cloud coding agents: every fresh sandbox has to be set up again with dependencies, credentials and project settings, which slows short tasks and makes long ones fragile. Making the environment a durable, shareable object brings Codex closer to how teams already manage development containers.

The review and security features push Codex from writing code toward checking it. Automatic first-pass reviews and scheduled repository scans mean the agent can generate work for humans to triage even when nobody asked it to, which puts more weight on how well it filters noise. OpenAI's emphasis on de-duplicating findings speaks to a familiar complaint about automated scanners that bury teams in repeated alerts.

Bundling Daybreak Blue into the scanner also widens access to OpenAI's security models. The company has been extending its Daybreak cyber tools to defenders, and this month extended access to Ukraine's government to help protect civilian infrastructure.

Context

The Bedrock integration is a notable concession to where many enterprises already run workloads. Letting customers keep OpenAI agents inside AWS spares companies that have standardized on Amazon from setting up a separate cloud relationship just to use them.

What to Watch

OpenAI has not said when Codex Security Cloud, the Decisions API or the Bedrock integration will be generally available, or what they will cost. Developers will also be watching how automatic reviews and scheduled scans are metered, since agent work that runs without a prompt can add up quickly.