Back to AI information
Codex Arrives on ChatGPT Mobile: Code Agents Start Coming Off the Desktop

Codex Arrives on ChatGPT Mobile: Code Agents Start Coming Off the Desktop

AI information Admin 131 views

On May 14, 2026, OpenAI announced that Codex has entered the ChatGPT mobile app preview, allowing iOS and Android users to view Codex's running tasks, approve commands, switch models, view terminal outputs, test results, screenshots, and diffs from their phones. Officially, Codex is currently used by more than 4 million users every week.

This is not a simple "mobile phone remote control computer". OpenAI's description is more like turning a long-running code agent into a cross-device workflow: Codex still runs on the user's laptop, development machine, or remote environment, files, credentials, and permissions remain on the original machine, and the mobile phone receives real-time status and context through a secure relay.

Why this matters

AI programming tools used to be tied to IDEs, terminals, or desktop applications, suitable for continuous work while sitting in front of a computer. Codex Mobile is a core value that allows users to make critical decisions while commuting, in between meetings, or away from their computers: answering clarifying questions, approving next steps, changing direction, and avoiding agents stopping while waiting for user input.

There are also new signals on the enterprise side. OpenAI also emphasized that Remote SSH is already available and Codex can enter managed remote environments; Hooks are officially available and can be used to scan for secrets in prompts, run validators, record sessions, or customize behavior by repository; Programmatic access tokens are available to Enterprise and Business plans.

Who to influence

Developers will feel the change first, especially those who let Codex run long tasks, fix bugs, refactor, and run tests. Enterprise teams should focus on approvals, credential retention, remote environment boundaries, and auditing capabilities. Mobile makes code agents more convenient, which also means that the risk of "approving a command at any time" is higher, and teams need to clarify which actions can be approved remotely and which must be confirmed back to a controlled development environment.

Source: OpenAI's official blog "Work with Codex from anywhere".

Recommended Tools

More