Allow ChatGPT to securely connect to active Codex sessions so users can monitor progress, review work, and send instructions through natural conversation.
Problem: When I use Codex, I often want to interact with it through ChatGPT instead of constantly switching between terminals, logs, and the Codex interface. Proposed feature: Allow ChatGPT to securely connect to one or more active Codex sessions with explicit user authorization. Once connected, ChatGPT could view tasks, monitor progress, read logs, review generated code, and send follow-up instructions or pause/stop/reprioritizE work.
Example workflow:
User: What is Codex working on right now? ChatGPT: Codex is implementing the authentication service. It has completed 12 files and is currently running tests.
User: Tell Codex to pause the refactor and fix the failing tests first.
ChatGPT: Instruction sent. Codex has paused the refactor and is prioritizing the failing tests.
Why this matters: This creates a true AI orchestrator experience where ChatGPT acts as the conversational coordinator while Codex focuses on execution.
Developers can manage long-running coding tasks entirely through natural language without constantly switching tools. All access should require explicit user permission, and users should always know which Codex session ChatGPT is connected to.