OpenAI Gives Codex Reusable Cloud Workspaces for Teams
OpenAI introduced reusable cloud development environments for Codex, its software engineering agent, at its Dev Day on Tuesday. The company described them as persistent, shareable workspaces that teams can open from any device, instead of temporary sandboxes created fresh for each task. TechCrunch
The event was DevDay 2026 in San Francisco on September 29. The environment update was part of a broader Codex and platform release covering the command line interface, code review, security scanning and agent APIs.
The reusable environments are meant to start tasks faster and give teams a shared workspace with approved settings and permissions. Dependencies, tooling, limits on credentials, and policy guardrails do not need to be rebuilt each run. State carries over. Teams start from the same baseline.
Cross-device access is central to the pitch. Codex is now available in the ChatGPT mobile app, after its earlier launch as a dedicated Codex app built to manage multiple agents at once. OpenAI has also documented a setup where a ChatGPT account lets users move between a local environment and Codex cloud without losing state.
Reusable environments and everyday automation
The focus on reuse builds on earlier internal work OpenAI described using Runme and WebMCP with Codex to turn repeated engineering tasks into workflows that can be reviewed and reused. The same approach applies to cloud migrations, where Codex can generate starter code and configuration such as Dockerfiles for cloud environments.
Temporary agents, known as ephemeral agents, are cheap to start and simple to understand. They do not retain organizational context well. Persistent environments reverse that tradeoff. They cut setup work but require ongoing management, access reviews, and cleanup of old state.
CLI, delegation and review
OpenAI announced a refreshed Codex CLI alongside the reusable cloud environments. The updated CLI lets developers start and direct tasks with their voice. That changes the input method. Dictation starts the task, and the agent still runs inside the configured environment.
OpenAI added a new /agents view in Codex to simplify delegation and tracking of multiple parallel tasks. The view is an operations screen for running many agents at once. Developers can launch concurrent agents, watch progress, and step in without juggling separate terminals or tabs.
OpenAI introduced a new Codex code review experience inside the ChatGPT desktop app for GitHub pull requests and GitLab merge requests. The review tools sit where developers already check code differences. It supports both hosting services without requiring a switch to the browser.
OpenAI's automatic code reviews let Codex do an initial review while users are away from their computers. The agent scans the change, leaves findings, and prepares follow-ups for human review on return. TechCrunch
Security Cloud
OpenAI launched Codex Security Cloud tools to help developers secure infrastructure. The system can scan full GitHub repositories on demand, on a regular schedule, and when new commits arrive. Those triggers match how security teams work. Manual audit, continuous checks, and checks tied to new code.
After scanning, Codex investigates findings, removes duplicates, and prepares fixes in the cloud. Sorting duplicates and preparing fixes is the slow step in most application security pipelines. Doing that triage in the same cloud environment where code runs shortens the path from alert to proposed patch.
Codex Security Cloud includes access to models from OpenAI's Daybreak Blue cybersecurity initiative without a separate application. Without a separate enrollment step, existing Codex users get the specialized models as part of normal use. TechCrunch
Decisions API and Agents API
OpenAI announced a new Decisions API for real-time decision-making that uses Luna to answer user-defined questions with predefined answers. The design limits output for speed-sensitive uses. Developers set the questions and the allowed answers. Luna returns a choice, useful for routing, gating, and policy enforcement where open-ended text would create risk.
OpenAI updated its Agents API to support computer use and the ability to use Amazon's Bedrock Managed Agents to build OpenAI agents that run entirely on AWS. Computer use adds mouse, keyboard, and interface navigation to the agent loop. Bedrock Managed Agents support lets teams keep runtime, data, and scaling inside AWS while still using OpenAI agent interfaces. TechCrunch
The broader context here is persistence plus governance. Reusable environments, approved settings, scheduled security scans, and predefined decision outputs point to agents that run longer, touch more systems, and need tighter operational controls. That is a different task from showing code completion. It is fleet management for autonomous work.
In my view, the most durable shift is not voice input or mobile access. It is treating the agent workspace as team infrastructure. I watched my own two kids pick up new tools at once and drop them just as fast when setup became tedious. Enterprise teams act much the same at scale. Persistent, pre-approved environments lower the effort needed for the tenth and hundredth task, which is where automation either builds on itself or stalls. If OpenAI can keep setup fast while keeping permissions auditable, Codex gets closer to a standard layer for software work rather than another side assistant.
