Originally published at devtoolpicks.com
OpenAI launched Codex in the ChatGPT mobile app yesterday, and the headline is doing more work than the feature. Let me give you the actual picture before you update your apps expecting something it is not.
Codex is now in preview on iOS and Android. It is available to all Codex users, including the free and Go plans. To use it, you update the ChatGPT mobile app and the Codex desktop app. Mac only for now, Windows is coming.
That is the straightforward part. Here is what it actually does.
What Codex Mobile Is (and Is Not)
This is not coding on your phone. Your MacBook, Mac Mini, or cloud devbox continues to run Codex entirely. The mobile app is a monitoring and control surface that connects to your desktop machine through a secure relay layer.
From the phone, you can:
- See the live state of your Codex environment
- Review outputs, screenshots, and status notes Codex sends as it works
- Approve commands that require sign-off before Codex proceeds
- Steer the task mid-execution (redirect, add context, correct course)
- Change which model is handling the current task
- Start new tasks from scratch
OpenAI describes the feature as going beyond remote control of a single task: from the phone, users can manage all threads, approve commands, change models, and start entirely new work.
The relay layer keeps your machine reachable without exposing it to the public internet. Active session state stays synced across devices through your ChatGPT account.
When This Actually Matters for Indie Hackers
Most indie hackers run focused interactive coding sessions. You open Claude Code or Codex, you iterate, you ship. For that workflow, mobile monitoring adds nothing because you are already at your desk.
The feature earns its keep in one specific scenario: long-running async tasks you set up and walk away from. A two-hour refactor. An overnight test suite run. A batch migration. A deployment pipeline that needs human approval at a checkpoint. This is where not having mobile visibility meant either staying at your desk or coming back to discover the agent stopped and waited for input three hours ago.
With mobile monitoring, you can check progress from the gym, get the mid-task approval notification while making dinner, or steer the agent during a commute instead of finding it idle when you get back.
If your Codex usage is primarily short interactive sessions, the honest answer is you will not notice this feature for weeks.
One More Caveat: Anthropic Got There First
Claude Code already has Remote Control, a mobile monitoring feature that lets you check on running sessions and approve commands from your phone. OpenAI catching up here is meaningful for Codex users, but it is not a first in the category.
The practical comparison:
| Feature | Codex Mobile | Claude Code Remote Control |
|---|---|---|
| iOS | ✅ | ✅ |
| Android | ✅ | Limited |
| Windows desktop support | Coming soon | Limited |
| Switch models from phone | ✅ | No |
| Start new tasks from phone | ✅ | No |
Codex mobile goes slightly further in what you can initiate from the phone. Claude Code Remote Control launched first. Neither is a decisive advantage if you are choosing an agent tool. The underlying coding quality matters more.
Our Codex vs Claude Code comparison covers the full capability picture if you are deciding between the two. The mobile feature does not change that verdict.
How to Get It
- Update the ChatGPT app on iOS or Android to the latest version
- Update the Codex desktop app on your Mac to the latest version
- Open the ChatGPT mobile app and look for the Codex section
- Your desktop environment should appear if both apps are updated and you are signed into the same ChatGPT account
The relay connects automatically. You do not need to configure anything beyond the app updates.
Note that during the preview period, access may queue or slow down under high demand. OpenAI is rolling this out gradually.
What This Means for the AI Coding Tool Race
OpenAI announced this the same week Anthropic changed how Claude subscriptions handle programmatic usage and the same week a critical Next.js vulnerability put self-hosted apps in the news. The pace of releases from both companies has been relentless since April.
Codex has 4 million weekly users according to OpenAI. Claude Code continues to grow in enterprise adoption. Both are adding features that make async multi-agent workflows more viable, which is where the real productivity gains for indie hackers will come from over the next twelve months.
If you are paying for a ChatGPT or Codex plan already, update your apps today and try it the next time you kick off a long background task. If you are evaluating which agent to invest in, the /goal command comparison is still the most relevant benchmark for how Codex and Claude Code handle complex agentic work.
The mobile app is a nice addition. The underlying agent quality is what actually determines whether you ship faster.
Top comments (0)