Loading video player...
OpenAI just added Sub Agents to the Codex App and Codex CLI for GPT-5.4, and this walkthrough breaks down how they work in practice. In this video, I cover: what subagents are why they matter for real coding workflows how Codex handles orchestration how approvals and sandboxing carry across agents how to define custom roles how to split work across explorer, reviewer, and implementation agents This is a practical overview focused on using Codex subagents for parallel software tasks, not just a surface-level product tour. *subscribe for more* Chapters 0:00 Why Subagents Matter 0:19 Availability Basics 0:34 How Coordination Works 0:50 Steering Active Work 1:05 Approvals And Sandboxing 1:20 Custom Roles 1:36 Role Splitting Strategy