OpenAI Launches Codex Mobile Feature, Expanding AI Programming to New Platforms
On May 15, 2026, AI research company OpenAI officially announced a mobile connection feature for its AI programming assistant, Codex. This function allows developers to pair their phone or tablet with a desktop environment running Codex, enabling remote code generation and interaction. This move is seen as a significant step into the mobile AI application market, aiming to meet developers’ programming needs in non-traditional work settings (like during commutes or travel) and enhance development flexibility and efficiency. Previously, third-party tools like Openclaw Hermes had already pioneered similar mobile solutions, gaining some market attention.
“Waiting for Desktop”: A Common Connection Bottleneck
Following the new feature’s release, numerous users on social media and tech forums reported encountering obstacles during the connection process. Specifically, after scanning the QR code generated on the desktop, the mobile device’s screen would remain stuck on “waiting for desktop,” preventing them from proceeding to the authorization step. Technical analysis suggests this problem is likely due to network complexities, particularly those related to network proxy configurations. When a user’s desktop computer uses a proxy server to access the internet, the Codex remote control service may fail to communicate correctly through that proxy, thus preventing the mobile and desktop clients from establishing a connection.
Forcing a Proxy via Command Line: An Effective Solution
To address this connection issue, the community has discovered a command-line-based solution. The core idea is to manually specify a proxy server for the Codex remote control process, forcing network traffic through the local proxy to bypass potential connection barriers. Here are the steps:
On the desktop computer where Codex is installed and running, open your system’s terminal (e.g., Terminal on macOS or Command Prompt on Windows).
In the terminal, enter the following environment variable configuration commands, followed by the command to run the Codex remote control program. Note that http://127.0.0.1:7890 is a common address and port for proxy services; you’ll need to replace it with your own proxy tool’s configuration.
HTTPS_PROXY=http://127.0.0.1:7890
HTTP_PROXY=http://127.0.0.1:7890
ALL_PROXY=http://127.0.0.1:7890
codex remote-control
After executing the command, you don’t need to wait for a specific output in the terminal. At this point, completely close the related application on your mobile phone.
Use your phone’s native camera app to rescan the new QR code generated on the desktop. You should now be correctly redirected to the authorization page. Once authenticated, the connection will be successful.
Open-Source Alternative: yepanywhere’s Multi-Model Support
Besides the official solution, an open-source GitHub project called yepanywhere offers another option for developers who need to manage multiple AI programming tools. According to its project description, yepanywhere uses underlying principles similar to Codex’s remote connection but boasts broader compatibility. It not only supports Codex but can also connect to and manage various AI models or IDE integrations, including Claude and Opencode, providing a unified mobile gateway for users with complex development environments.
Impact of the Official Launch on the Third-Party Tool Ecosystem
The official launch of Codex’s mobile connection has sparked discussions about the future of third-party tools like Openclaw Hermes. While OpenAI’s official support may offer superior performance and stability, third-party tools might still retain their value in niche markets through deep integrations with specific platforms (like WeChat Work and Feishu) and other differentiating features. In the future, the interaction and competition between the official offering and the community ecosystem will collectively drive the evolution of AI-assisted programming tools.
