Feature request: identify the agent controlling Computer Use
Nobody has claimed this yet.
- Dominant language
- Rust
- Stars
- 125k
- Forks
- 19.5k
- PR merge metrics
- PR metrics pending
Description
What version of the Codex App are you using?
Feature request; the behavior was observed in the Codex Desktop app.
What platform is your computer?
Windows
What issue are you seeing?
When Computer Use is active, the UI shows an ESC hint to cancel the current action, but it does not make it clear which agent or chat currently controls the computer.
This is particularly ambiguous in multi-agent workflows, where several agents can be active at the same time.
What is the expected behavior?
Please identify the active Computer Use owner in the same visible area as the cancellation control.
For example:
ESC to cancel · Agent: Fix login flow
Possible approaches:
- Show the active agent or chat name in the Computer Use window header.
- Show that agent's avatar, or a compact status icon, next to the cancellation hint.
- Use a distinct visual indicator when control moves to another agent, and update it immediately on handoff.
The indicator should make it obvious which task is acting without requiring the user to inspect each chat.
Why this matters
This improves transparency and supervision of concurrent agent activity, and reduces the chance of interrupting the wrong task.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Review the Computer Use window and its existing ESC cancellation hint in the Codex Desktop app. Trace how the active agent or chat and handoffs are represented, then determine how to show the owner alongside the cancellation control. Done means concurrent tasks are distinguishable, the indicator updates on handoff, and users can identify the controlling agent without inspecting each chat.
Written by the indexing model from the issue text.
Assessment
- Domain
- design, desktop
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Quiet
- Clarity
- Mostly clear
- Newbie friendliness
- 45/100