The Codex mobile app is now available in preview within the ChatGPT mobile application, enabling users to manage development work from anywhere. The feature rolls out across iOS and Android for all subscription plans, including Free and Go tiers, in all supported regions.
More than 4 million people use Codex every week, according to the announcement. The mobile integration addresses the emerging need for asynchronous collaboration as agents handle longer-running tasks. Users can now answer questions, review findings, change direction, approve commands, or contribute new ideas directly from their phones while Codex operates across laptops, development boxes, or remote environments.
Codex Mobile App Features and Functionality
The Codex mobile app delivers a fully-featured experience for remote work management. When connected to any machine where Codex is running, the app loads the live state from that environment. Users can work across active threads, approvals, plugins, and project context without being limited to single-task remote control.
From a phone, users can review outputs, approve commands, change models, or initiate new work. Files, credentials, permissions, and local setup remain on the machine where Codex operates, while updates flow back to the phone in real time. This includes screenshots, terminal output, diffs, test results, and approvals. A secure relay layer keeps trusted machines reachable across devices without exposing them directly to the public internet.
Real-World Use Cases for Mobile Codex
The release highlights several practical scenarios where mobile access improves workflow efficiency. Users can start investigating a bug while waiting for coffee, with Codex inspecting relevant files, reproducing issues in the browser, and running tests from the development environment. If Codex requires clarification or permission, users can respond or approve from their phone and follow along with live screenshots and test results.
During commutes, users can reach decision points on longer tasks. If Codex discovers multiple viable approaches and needs direction, users can review tradeoffs and choose a path from their phone, keeping work moving in the intended direction. In customer-facing scenarios, users can ask Codex to synthesize updates from Slack, email, and documents before important calls, and refresh summaries if new details emerge.
Enterprise and Remote Environment Support
For teams using managed remote environments, Remote SSH is now generally available. Codex can connect directly into these environments, with the desktop app automatically detecting hosts from SSH configuration. Once connected, remote environments become accessible across authorized ChatGPT devices through the same secure relay infrastructure.
Several updates expand team capabilities at scale. Programmatic access tokens, issued directly from ChatGPT workspace settings, provide scoped credentials for CI pipelines, release workflows, and internal automations. Hooks are now generally available and can scan prompts for secrets, run validators, log conversations, create memories, or customize Codex behavior for specific repositories and directories.
Support for HIPAA-compliant use of Codex is now available in local environments (CLI, IDE, App) for ChatGPT Enterprise workspaces. This enables healthcare organizations to support patient care and operational workflows. The Codex mobile app is rolling out in preview on iOS and Android across all plans. Support for connecting phones to the Codex app on Windows is coming soon.
Integration with Artificial Intelligence Workflows
Codex functions as an artificial intelligence agent capable of handling extended work tasks while maintaining human oversight. The mobile app preserves the collaborative rhythm required when agents take on longer-running projects. Real-time synchronization across devices keeps users connected to active work without requiring constant attention at a single machine.
The expansion to mobile represents a significant shift in how development teams interact with AI-powered tools. By allowing intervention at critical decision points, the system balances automation with human judgment. This model supports faster iteration cycles and reduces the risk of unnecessary rework caused by unclear context or outdated information.
