Mobile companion sync for desktop and cloud tasks #20757

Codex currently lacks a first-class mobile companion experience that can sync with the Codex desktop app and Codex cloud tasks. This is a serious product gap for anyone using Codex as a long-running agentic development command center.

Codex is explicitly designed around long-running tasks, parallel agents, background work, worktrees, cloud tasks, automations, and human supervision. But right now, when I leave my computer, I lose the practical ability to monitor progress, respond quickly, approve or steer work, and start new Codex work from my phone.

This is not a minor convenience feature. It is a core workflow requirement.

Current problem

Today, Codex desktop work is effectively tied to the desktop environment. The available mobile experience through ChatGPT/Codex is not enough to act as a proper companion to the desktop Codex app.

The missing capabilities are:

• View active Codex desktop threads from mobile.
• See real-time progress for local, worktree, and cloud tasks.
• Receive mobile notifications when Codex needs input, approval, or review.
• Reply to an active desktop Codex thread from mobile.
• Start a new task from mobile and assign it to a desktop project, cloud environment, or known repo.
• Review task summaries, diffs, files changed, and test results from mobile.
• Decide whether to continue locally, move to cloud, create a PR, or pause a task.
• Manage automations and long-running follow-up loops from mobile.

Without this, Codex is not truly usable as an always-on agent command center. It becomes a desktop-only control panel for workflows that are explicitly meant to continue over time.

Why this matters

Agentic coding is not a short chat interaction. Codex tasks can run for minutes, hours, or longer. Developers frequently step away from their computer while tasks are running. In those moments, the ability to supervise from mobile is essential.

A mobile companion experience would let users:

• Check whether a task is blocked.
• Approve or reject requested actions.
• Give additional instructions before the agent goes too far in the wrong direction.
• Start new work immediately when an idea, bug report, CI failure, or production issue appears.
• Keep long-running development work moving without needing to reopen the desktop app.
• Use Codex more like a real asynchronous engineering teammate.

This directly affects the usefulness, reliability, and adoption of Codex for serious engineering workflows.

Expected solution

OpenAI should provide an official mobile companion experience for Codex, either inside the ChatGPT mobile app or as a dedicated Codex mobile app.

At minimum, it should support:

1. Cross-device thread visibility
   Mobile should show active and recent Codex desktop/cloud threads under the same ChatGPT account.

2. Real-time progress monitoring
   Users should be able to see status, current step, changed files, test results, blockers, and final summaries.

3. Mobile interaction with active tasks
   Users should be able to send follow-up prompts, answer questions, approve/deny actions where safe, and steer running work.

4. New task creation from mobile
   Users should be able to start new Codex tasks from mobile and target a known repo, project, cloud environment, or desktop project.

5. Notifications
   Codex should send push notifications for blocked tasks, completed runs, requested approvals, failed tests, automation findings, and PR-ready results.

6. Review support
   Users should be able to inspect diffs, summaries, logs, and PR creation options from mobile.

7. Clear security boundaries
   For local desktop tasks, mobile actions can require explicit preconfigured trust settings. This should be designed carefully, but it should not be used as a reason to leave the workflow unsupported.

Security considerations

This feature obviously needs careful permissioning. But the right answer is not to avoid the capability. The right answer is to design it with clear controls:

• Per-device authorization.
• Per-project mobile access settings.
• Separate permissions for read-only monitoring, commenting, approvals, and task creation.
• Optional restriction to cloud tasks only.
• Audit logs for mobile-issued commands.
• Ability to disable mobile control in enterprise workspaces.

Request

Please treat this as a high-priority product gap.

Codex is being positioned as the central command center for agentic software development. A command center that only works when I am physically sitting at my computer is incomplete.

Mobile supervision and mobile task creation are essential for the long-running, asynchronous workflows Codex is built around. OpenAI should prioritize an official Codex mobile companion experience as soon as possible.