OpenAI has integrated Codex into the ChatGPT mobile app on iOS and Android, available in preview across all plan levels. Developers can now monitor live environments, approve pending commands, and switch between threads without touching a laptop. Another move in the agentic coding race against Anthropic and Claude Code.
Key Takeaways
- Codex is now accessible on iOS and Android in preview, available on all ChatGPT plans
- Live environment monitoring, command approval, and multi-thread management from mobile
- Anthropic launched a similar remote control feature for Claude Code back in February 2026
What Codex Can Do From Your Phone
Until now, Codex was a desktop tool. OpenAI’s coding agent ran in cloud environments, but tracking and control stayed tied to a laptop screen. As of May 14, 2026, that constraint is gone.
OpenAI has embedded Codex directly into the ChatGPT mobile app on iOS and Android. The feature is rolling out in preview for all plan tiers, with no additional cost announced at this stage. Developers can view live Codex environments, manage multiple threads, review outputs, approve commands, switch models, and start new tasks.
OpenAI’s own framing: “From your phone, you can work across all of your threads, review outputs, approve commands, change models, or start something new.”
This is not a read-only dashboard. It is a full mobile control panel. A developer can launch a task from their desk, step into a meeting, and from their phone check progress, adjust course, or unblock a command waiting for approval. The workflow no longer stops when the laptop closes.
This release fits into a rapid sequence. In April, OpenAI shipped background task capabilities for Codex on desktop. In May, a Chrome extension for browser sessions. The mobile version completes this infrastructure of permanent access to the agent across devices.
OpenAI Pushes Hard, Anthropic Too
The timing is not accidental. Anthropic had already launched an equivalent remote control feature for Claude Code in February 2026, several months before this mobile Codex release. As we covered when AI coding costs started alarming engineering budget owners, Claude Code and Codex are competing for the same territory: technical teams offloading more and more tasks to autonomous agents.
In this context, every feature release is as much a commercial argument as a technical one. Offering mobile access to a coding agent reduces friction for distributed teams, developers on the move, and technical leads who supervise without coding themselves.
The competition between OpenAI and Anthropic in agentic coding is not only a benchmark war. It is a war of integration into daily work habits. Every additional touchpoint, every friction reduced, builds dependency and makes switching tools more costly for the user.
OpenAI is leveraging the depth of the ChatGPT ecosystem. The mobile app is already installed on hundreds of millions of phones. Distributing Codex through that existing app is an adoption shortcut few competitors can replicate.
Also on Horizon:
- Alexa for Shopping: Amazon Brings AI Into Its Search Bar
- Sam Altman on the Stand: The Question of Trust
- Ramp AI Index: Anthropic Edges Out OpenAI in Enterprise
Agentic Coding Leaves the Office
The real shift this announcement signals is broader than the feature itself. Agentic coding is gradually leaving the developer’s desk to become a distributed capability, accessible from any device at any time.
This raises questions about roles within technical teams. If an engineering lead can supervise multiple coding agents from a phone during a commute, the scope of what is “delegable” to the machine expands mechanically. The question is no longer whether AI can code, but how many tasks a single developer can oversee in parallel.
In the short term, teams adopting Codex mobile will gain continuity: agents no longer idle while waiting for a developer to sit back down at a screen. Development cycles can compress on lightly supervised tasks.
Over the medium term, the proliferation of mobile coding agents will accelerate conversations around governance and cost control. OpenAI anticipated this with the launch of Daybreak, its AI cybersecurity platform targeting environments where code is generated and executed autonomously at scale. The more pervasive agentic coding becomes, the larger the attack surface grows.
The feature is in preview. Final usage limits and pricing have not been announced. But the direction is clear: the coding agent now follows its user everywhere, and both major players in this market are competing on that front.
Follow the story on Horizon.
