Subagents (Codex) is a workflow feature in 📝OpenAI's Codex platform that enables spawning parallel AI agent sessions from a parent Codex session to handle complex tasks concurrently. Each subagent is configured via TOML files that define parameters including model selection, sandbox mode, nickname candidates, MCP servers, and skills. The architecture mirrors 📝Claude Code's explorer/worker/default subagent pattern and allows developers to orchestrate large numbers of specialized agents for parallel execution of tasks such as codebase exploration, PR review, or multi-step feature implementation.
