OpenAI brings Codex to the ChatGPT mobile app in preview

news

OpenAI says Codex can now be used from the ChatGPT mobile app in preview, letting users start work, review outputs, and approve next steps from iPhone or Android while execution stays on their machines.

OpenAI Developers remote connections image for Codex

What happened

OpenAI has started rolling out Codex inside the ChatGPT mobile app in preview on iOS and Android. The company says users can now start new work, review outputs, answer questions, approve next steps, and steer active threads from their phones while Codex continues running on a connected laptop, Mac mini, devbox, or other remote environment.

This is not just a notification layer. OpenAI is positioning the mobile app as a live control surface for longer-running Codex sessions, which matters because more agent-style development work now runs for long enough that people need to check in, unblock it, or redirect it away from their desk.

What the official source confirms

OpenAI’s official product post, Work with Codex from anywhere, confirms that Codex in the ChatGPT mobile app is now in preview and is rolling out across all plans, including Free and Go, in supported regions. According to OpenAI, the mobile experience can load the live state from machines where Codex is already running, including active threads, approvals, plugins, screenshots, terminal output, diffs, and test results.

The same announcement also confirms that Remote SSH is now generally available, so teams can connect Codex directly to managed remote environments, and that Hooks are now generally available for more customization and automation. OpenAI’s developer documentation around Remote connections further explains the model: the phone acts as a trusted companion interface, while files, credentials, permissions, and execution stay on the host machine rather than moving onto the handset itself.

Why the story is trending on X

The story is moving on X because the launch turns Codex from a desktop-first coding agent into something closer to an always-with-you workflow. OpenAI’s main account posted the launch with a clearer mainstream framing, saying users can start work, review outputs, steer execution, and approve next steps from the ChatGPT mobile app. That post quickly picked up strong traction, reaching roughly 4.48 million views, 21.7K likes, and 2.6K reposts at the time of writing.

The developer-side announcement from @OpenAIDevs also gained solid engagement, with roughly 1.28 million views and thousands of interactions. That combination matters: the consumer-facing account pushed the bigger product narrative, while the developer account signaled that this is meant for real coding workflows rather than a lightweight chat companion feature.

What this means for developers, builders, and product teams

For developers, this is a meaningful workflow shift. Long-running coding agents become more useful when approvals and course corrections are not trapped behind one machine. If a task hits a decision point during a commute, a meeting, or while away from the desk, the work no longer has to stall.

For product teams, the more interesting signal is that OpenAI is treating continuity across devices as part of the agent product itself. The hard part is not only model quality. It is keeping the loop alive across trust boundaries, permissions, context, and user attention. OpenAI’s relay-based approach suggests the next phase of agent products will compete on orchestration and session continuity just as much as raw coding performance.

For companies building developer tools, this raises the bar. A coding agent that only works well when someone is sitting in front of one terminal now looks more limited. Users will increasingly expect background execution, mobile approvals, remote state sync, and secure control over multiple environments.

What remains unclear

A few important questions are still open. OpenAI says the feature is rolling out in preview, so real-world stability, latency, and reliability across different host setups will take time to judge. The company also has not said in detail how often mobile usage will still depend on the desktop Codex app versus other Codex entry points over time.

It is also not fully clear how deeply teams will adopt this outside early power users. The feature is easy to understand, but the habit change is bigger than a normal mobile companion app because it asks developers to trust longer autonomous sessions and intervene only when needed.

Sources