Codex

Codex is OpenAI’s coding agent for development work such as reading codebases, proposing edits, and operating through terminal or editor in multi-step loops. The name was also historically used for OpenAI’s early code models, so the term appears both as a product and as a reference to that model lineage.

For editor and terminal workflows, compare agentic coding, ChatGPT, Cursor, and Claude Code. For review-focused workflows, tools like CodeRabbit are often used.


Key characteristics

  • Is focused on coding workflows where the model does more than autocomplete and can participate in full development loops.
  • Extends the idea of early code models into more agentic workflows with tools and feedback.
  • Works best when combined with tests, review, and clear project constraints.