Codex multi-agent orchestration sucks

Open 💬 5 comments Opened May 20, 2026 by OpEnD17

What variant of Codex are you using?

App

What feature would you like to see?

Codex barely uses sub-agents properly in long tasks.

When analyzing large execution traces or workflows with multiple agent jobs, the main agent usually reads everything itself instead of delegating work to sub-agents.

Additional information

_No response_

View original on GitHub ↗

This issue has 5 comments on GitHub. Read the full discussion on GitHub ↗