Allow Codex Desktop Agent to interact with integrated terminal sessions

What variant of Codex are you using?

App

What feature would you like to see?

Feature Request

Allow Codex Desktop Agent to send input to existing integrated terminal sessions.

Scenario

I use the Codex Desktop App workspace with integrated terminal panels running other coding agents such as Claude Code and Kimi Code.

Current limitation

Codex can read and observe terminal output, but it cannot send input to an already running interactive CLI session.

Problem

When another CLI agent requires input, confirmation, or additional instructions, I have to manually switch to the terminal panel and type commands. This interrupts the multi-agent workflow.

Suggested improvement

Allow Codex Desktop Agent to interact with existing terminal sessions with explicit user permission:

• Select an existing terminal session
• Send text input to the process stdin
• Ask for confirmation before sending commands

Use case

Codex could work as an orchestrator:

Codex plans → Claude Code implements → Kimi Code reviews → Codex verifies.

This would make Codex Desktop a stronger multi-agent development environment.

Additional information

This feature should be opt-in and require explicit user confirmation before sending any input to another terminal process.

The goal is not autonomous control of other agents, but enabling a safe multi-agent workflow where users can coordinate different coding agents inside the same Codex workspace.

An integrated terminal interaction layer would make Codex Desktop a more powerful environment for modern AI-assisted software development.

Start with the Codex Desktop App's integrated terminal panels and trace how existing terminal sessions are selected and observed. Define how explicit confirmation would govern sending text to another process's stdin, then verify the workflow with an existing interactive CLI session and confirm that input is never sent without permission.