OpenAI expanded its Codex coding agent to the ChatGPT mobile app on May 15, 2026, enabling iPhone and Android users to remotely monitor and manage coding sessions running on their computers. The integration allows users to keep track of tasks from mobile devices while Codex continues executing workflows across laptops, dedicated development boxes, or other remote environments.

To activate the feature, users must update both their desktop Codex application and the ChatGPT mobile app. Pairing is completed by selecting the new Codex mobile section within the desktop interface and scanning a generated QR code with a phone. Once linked, the mobile app mirrors the live state of the desktop session, providing users with the ability to initiate new projects, review outputs, approve commands, and switch between models in real time. Notifications are delivered to the phone when the agent completes a specific task or requires manual input.

OpenAI confirmed that sensitive data, including local files, credentials, and access permissions, remains stored exclusively on the host computer where the agent operates. Only non-sensitive information such as terminal outputs, screenshots, test results, diffs, and approval requests are transmitted to the mobile device. The company advised users to only pair devices they personally own and trust to maintain security protocols. While the tool is currently available in preview on mobile, OpenAI noted that support for remote access to Codex sessions running on Windows environments is expected to be released in the near future.

This mobile integration builds upon the recent launch of a Codex Chrome extension, which enables the coding agent to function directly within web browsers to pull context from open tabs and test web applications. The feature is part of a broader expansion of OpenAI's developer-focused tools, following the recent introduction of the GPT-5.5 model, which the company claims provides enhanced capabilities for complex multi-step planning, tool utilization, and autonomous verification of output during agentic coding and scientific research projects.

OpenAI expanded its Codex coding agent to the ChatGPT mobile app on May 15, 2026, enabling iPhone and Android users to remotely monitor and manage coding sessions running on their computers. The integration allows users to keep track of tasks from mobile devices while Codex continues executing workflows across laptops, dedicated development boxes, or other remote environments.

To activate the feature, users must update both their desktop Codex application and the ChatGPT mobile app. Pairing is completed by selecting the new Codex mobile section within the desktop interface and scanning a generated QR code with a phone. Once linked, the mobile app mirrors the live state of the desktop session, providing users with the ability to initiate new projects, review outputs, approve commands, and switch between models in real time. Notifications are delivered to the phone when the agent completes a specific task or requires manual input.

OpenAI confirmed that sensitive data, including local files, credentials, and access permissions, remains stored exclusively on the host computer where the agent operates. Only non-sensitive information such as terminal outputs, screenshots, test results, diffs, and approval requests are transmitted to the mobile device. The company advised users to only pair devices they personally own and trust to maintain security protocols. While the tool is currently available in preview on mobile, OpenAI noted that support for remote access to Codex sessions running on Windows environments is expected to be released in the near future.

This mobile integration builds upon the recent launch of a Codex Chrome extension, which enables the coding agent to function directly within web browsers to pull context from open tabs and test web applications. The feature is part of a broader expansion of OpenAI's developer-focused tools, following the recent introduction of the GPT-5.5 model, which the company claims provides enhanced capabilities for complex multi-step planning, tool utilization, and autonomous verification of output during agentic coding and scientific research projects.