OpenAI Codex Subagents Now Generally Available for Parallel Agent Workflows

OpenAI has announced general availability of subagent workflows in Codex, after several weeks in preview behind a feature flag.

Codex can now spawn specialized agents in parallel to handle complex, multi-step tasks. Users define custom agents with different model configurations and instructions, and Codex orchestrates them to work concurrently. This is particularly useful for codebase exploration, multi-file refactoring, or implementing feature plans that span multiple components.

Subagents only activate when explicitly requested. Each subagent runs with its own context and instructions, and results are collected into a single coordinated response. Subagent activity is currently surfaced in both the Codex app and CLI, with IDE extension visibility coming soon.

The GA release enables subagent workflows by default in current Codex releases, removing the need for feature flag opt-in.

https://developers.openai.com/codex/subagents