[Feature Request] Introduce a status indicator to prevent "Waiting for agents" terminal spam #33485

Activity

1. B143KC47

   added

   enhancementNew feature or requestNew feature or request

   on Jul 16, 2026
2. github-actions

   added

   CLIIssues related to the Codex CLIIssues related to the Codex CLI

   subagentIssues involving subagents or multi-agent featuresIssues involving subagents or multi-agent features

   TUIIssues related to the terminal user interface: text input, menus and dialogs, and terminal displayIssues related to the terminal user interface: text input, menus and dialogs, and terminal display

   on Jul 16, 2026
3. criticaldeveloper commented on Jul 22, 2026

   

   criticaldeveloper

   on Jul 22, 2026

   More actions

   Hi, I've found the same problem but totally mitigate it adding the following to the orchestrator instructions (you can basically copy/paste it to codex-cli so it swallows it into the current documentation):

   ## Subagent waiting policy
   
   - Never use `wait_agent` as a periodic heartbeat or tight polling loop.
   - Continue useful parent-thread work while subagents are running whenever possible.
   - Call `wait_agent` only at a genuine synchronization barrier.
   - When waiting, include all relevant active agent IDs in one call and use a long,
     bounded timeout rather than repeated short waits.
   - After `wait_agent` returns with no completed agents:
     - Do not immediately call `wait_agent` again.
     - Check agent state once.
     - Send at most one progress request to an apparently stalled agent.
     - Continue independent work, recover the task in the parent, or terminate the
       stalled agent.
   - Permit at most one consecutive no-progress wait.
   - Never enter an unbounded wait/retry loop.
   - Render the orchestration status table only when state changes.

   With that, the main orchestrator agent stops polling and it still receives the update automatically from the subagent when it finishes the whole task chain (so no polling is really needed).

   With that said, I wouldn't consider that a solution, because proper agent visualization must be a requirement in agentic development.

   ❤️React with ❤️3Reacted by KO Ho Tin, Johnny Bergström and Alan Ahn
4. WoojinAhn commented on Sep 21, 2026

   

   WoojinAhn

   on Sep 21, 2026 – with ChatGPT Codex Connector

   More actions

   +1. The key issue for me is discoverability: I want to notice immediately when subagents are running, without opening a separate view.

   In Codex CLI 0.155.1, subagent activity appears in the conversation history, and I can inspect agents through /subagents. However, the bottom of the main view does not provide a persistent overview of running subagents. This is especially noticeable when the parent finishes responding while background agents are still working.

   Claude Code’s CLI makes this easier to notice by displaying agent activity near the bottom.

   A compact indicator that appears automatically when subagents start would help—for example, 2 subagents running, expandable to show their names and statuses. It should remain visible while they are running, even after the parent’s turn ends.