What Is OpenAI’s Codex Cloud? The DevDay Upgrade Explained

Close-up of a developer's hands typing code on a MacBook with a code editor on screen

OpenAI’s coding agent just moved out of your laptop. At DevDay 2026 in San Francisco on September 29, OpenAI gave Codex — its software engineering agent — reusable cloud environments that persist across sessions and devices. Close the laptop, pick up your phone, and the work keeps going.

What Codex Cloud Actually Is

Each task you hand it runs inside an isolated container on OpenAI’s infrastructure, preloaded with a clone of your repository and its dependencies. It writes code, runs tests, and returns a diff and summary for your review.

Previously, those cloud tasks were one-off sandboxes: spun up, used once, thrown away. The DevDay upgrade makes the environments reusable and persistent — configured once, shared across your team, and reachable from a computer, a phone, or the web. According to TechCrunch, the point is faster task startup and a shared workspace with approved settings and permissions, rather than each developer’s quirks.

Everything Codex Got at DevDay 2026

The cloud environments were the headline, but OpenAI shipped a full bundle of Codex upgrades:

  • Reusable cloud environments — persistent, configurable workspaces that teams can share. Tasks can start from a computer or phone and keep running in the cloud.
  • Refreshed Codex CLI — the terminal app now takes voice commands to start and steer tasks, plus a new /agents view for tracking multiple tasks at once, with cleaner prompt editing, session resuming, and worktree handling.
  • Code Review in the desktop app — read summaries, explore diffs, and ask Codex about issues before posting feedback on GitHub pull requests or GitLab merge requests. Automatic first-pass reviews run while you’re away.
  • Codex Security Cloud — on-demand or scheduled scans of whole GitHub repos: findings investigated, duplicates removed, fixes prepared in the cloud — including OpenAI’s Daybreak Blue security models with no separate application.
  • Decisions API (limited preview) — a fast endpoint built on the Luna model that picks from answers you define, aimed at classification and routing.
  • Agents API with computer use (public beta) — agents can now operate software through its UI, and Amazon’s Bedrock Managed Agents run OpenAI agents entirely inside AWS.

Availability: Codex cloud runs on Plus and every tier above it. The refreshed CLI and Code Review reach all plans, according to OpenAI’s own DevDay recap.

Codex Cloud vs. the Other AI Coding Tools

Here’s the practical decision framework — where Codex Cloud fits, and when you’d pick something else:

Tool Where the work happens Best for Access
Codex Cloud OpenAI’s cloud, background Async batch jobs: features, bug fixes, test sweeps, repo reviews ChatGPT Plus and up
Codex CLI Your terminal, interactive Active real-time coding sessions Paid plans
Cursor Your IDE, interactive Active real-time coding sessions Paid plans
GitHub Copilot Inside your IDE Autocomplete and in-editor suggestions Paid plans

The key question isn’t which tool is “smartest” — it’s where your attention goes. Codex Cloud is built for delegation: you feed it a queue, it works alone, and your only job is review. The others keep you in the loop while the code gets written.

Why This Matters

The DevDay upgrade flips who does the waiting. Coding agents used to demand your machine stay awake and your attention stay engaged; now the agent becomes infrastructure you schedule, and attention is the scarce resource you spend only on review.

The honest version: this is OpenAI’s claim until teams test it on their own codebases. A shared, reusable environment centralizes trust — if its configuration or permissions are wrong, every task inherits the mistake. Start with low-risk maintenance work before handing it production migrations. But for teams drowning in “works on my machine” bugs, a team-approved starting point that survives the laptop lid closing is a genuine workflow change, not a demo trick.

Meanwhile in Apps & Tools:

Written by
Daniel covers apps, software, and productivity tools — from AI-powered apps and major platform updates to the tools teams actually use to get work done.