codex-worker

You can run multiple Codex agents in parallel by orchestrating them through isolated git worktrees and tmux sessions. This setup allows each agent to handle an independent subtask concurrently, reducing overall completion time for complex projects.

To spawn parallel Codex CLI workers, your environment must have the codex CLI, tmux, and a git repository configured to support worktrees. These preflight requirements ensure isolated workspaces and terminal sessions for each agent.

You should orchestrate Codex workers instead of the built-in Task tool when codex and tmux are available, enabling richer agent capabilities. It is ideal for decomposing tasks into independent subtasks like triaging issues or parallel module refactors.

Yes, you can manage and clean up parallel git worktrees and tmux sessions using standardized naming conventions and lifecycle commands. This approach simplifies monitoring active agents and safely removing worktrees after task completion.

The best way to refactor multiple modules concurrently is to decompose the task into independent subtasks and spawn a dedicated Codex worker for each. Each worker operates in its own isolated git worktree and tmux session to prevent conflicts.