Codex Guide

What it does

Codex is OpenAI's coding agent, delivered through a ChatGPT account rather than sold as a separate product. Give it an instruction and it plans the steps itself, reads and edits files, and runs the commands needed to get the job done. You can drive it from a terminal or from a dedicated app that shows the work as it happens.

The AI in this role does not just answer in a chat window: it runs in a terminal (a screen where you type commands to control the computer directly) and actually creates files and runs code. Once you give it an instruction, it chains several steps together on its own.

When it fits

It fits teams that already carry a paid ChatGPT plan at the organization level, since no separate contract is needed to start using it. If several people share it, plan the workload with each person's own usage limit in mind, since Codex draws down against the account it runs under.

When it doesn't fit

The limitations below come from the vendor's own official documentation, not our evaluation. As of 2026-09-22

Codex only runs under a ChatGPT account and follows that plan's rate limits, so it cannot be bought or used on its own. It is also reachable on the free ChatGPT and Go tiers, but OpenAI describes that free access as a temporary arrangement rather than a lasting free tier.

• View source at openai.com

Compare with

Claude Code sits closest to it. Codex fits if your team already lives in the OpenAI ecosystem and wants a visual app to watch progress in, and Claude Code fits if a conversational, multi-step handoff style suits how the team works.

• Claude Code

Official link

Other tools in this layer

• Claude Code
• Hermes Agent
• OpenClaw