Overview
Codex is the umbrella name for OpenAI’s agentic coding surfaces, a terminal CLI, an IDE extension, a cloud agent you delegate to from ChatGPT, and a GitHub bot, all sharing one underlying Codex model (the GPT‑5.x‑Codex line) and one account context. It reads your codebase, edits files, runs shell commands, and runs tests in a loop, and can sustain long, mostly autonomous sessions on real engineering tasks.
What you get
Agentic, project-level coding rather than single-snippet completion. The CLI exposes a structured /plan, /exec, /review loop; the cloud side runs in its own worktrees and environments so agents can work in parallel across projects. Skills and Automations let it pick up routine work (triage, CI chores, docs) on a schedule and drop results into a review queue, so you supervise outcomes instead of typing every edit.
Where it fits in your stack
Codex sits at the “hand off the whole task” end of the AI coding spectrum: you describe the goal, it produces a working project, you review the diff. Most people pair it with version control and a real review step, since an agent that runs hundreds of tool calls unattended needs guardrails on scope, secrets, and what it’s allowed to execute.
Who it’s for
Builders and teams who’d rather delegate a feature or a whole small app to an agent than steer each line. When evaluating, weigh how it behaves on your real repos, how it handles ambiguity, the quality of its self-testing, and whether the autonomy level matches how much oversight you want.
Pricing & licensing
Bundled with paid ChatGPT plans (Plus, Pro, Business, Enterprise) under usage limits, with API access to the Codex models for programmatic use. Pricing and limits change regularly, so check OpenAI’s Codex page for current details before you commit.