OpenAI showed off a set of upgrades for its software‑engineering agent Codex on Sep 29, 2026. The headline change is a reusable cloud development environment that developers can open on a laptop, phone or any web‑connected device. The environments stay alive between sessions, so work can pick up where it left off instead of starting a fresh sandbox each time.
Alongside the persistent clouds, OpenAI rolled out a refreshed Codex command‑line interface (CLI) that accepts voice commands, a new /agents view for tracking multiple tasks, and a cleaner terminal UI for longer sessions.
New code‑review and security features
Codex is now embedded in the ChatGPT desktop app for code reviews. Users can ask the agent to summarize changes, flag possible bugs, or generate feedback for GitHub pull requests and GitLab merge requests. An automatic review mode can run while the developer is away.
On the security side, OpenAI launched Codex Security Cloud. The tools can scan whole GitHub repos on demand or on a schedule, de‑duplicate findings, and suggest fixes—all done in the cloud, even when a laptop is closed. The feature also includes access to OpenAI’s Daybreak Blue cybersecurity models without a separate application.
API extensions and cloud agents
The company added a Decisions API that uses the Luna model to answer predefined questions in real time. It also updated the Agents API so agents can control a computer and run fully on Amazon Bedrock Managed Agents, letting developers build OpenAI agents that live entirely on AWS.
Why it matters
Developers can stop rebuilding their workspace each time they switch devices, saving time and reducing setup errors. The built‑in code review and security scans mean teams get faster feedback and fewer vulnerabilities, even when no one is at their computer.
