Overview

OpenAI Codex CLI is an open‑source command‑line tool that brings the power of our latest reasoning models directly to your terminal. It acts as a lightweight coding agent that can read, modify, and run code on your local machine to help you build features faster, squash bugs, and understand unfamiliar code. Because the CLI runs locally, your source code never leaves your environment unless you choose to share it.

For more details about Codex, please check out the Github repo directly - https://github.com/openai/codex.

Key Functionality

Zero‑setup installation – a single npm install -g @openai/codex (or codex --upgrade) gets you started.
Multimodal inputs – pass text, screenshots, or diagrams and let the agent generate or edit code accordingly.
Rich approvals workflow – choose how hands‑on you want to be with three distinct modes (see Approval Modes).
Runs entirely in your terminal – perfect for quick iteration without switching contexts.

Approval Modes

Mode	What the Agent Can Do	When to Use
Suggest (default)	Read files. Proposes edits & shell commands, but requires your approval before making changes or executing commands.	Safe exploration, code reviews, learning a codebase.
Auto Edit	Read and write files automatically. Still asks before running shell commands.	Refactoring or repetitive edits where you want to keep an eye on side‑effects.
Full Auto	Read, write, and execute commands autonomously inside a sandboxed, network‑disabled environment scoped to the current directory.	Longer tasks like fixing a broken build or prototyping features while you grab a coffee.

Pro tip: Codex warns you before entering Auto Edit or Full Auto if the directory isn’t under version control.

Quick Start

Install: npm install -g @openai/codex
Authenticate: Export your OpenAI API key (export OPENAI_API_KEY="<OAI_KEY>").
Run in Suggest mode: From your project root, type codex and ask for example: “Explain this repo to me.”
Switch modes: Add flags as needed:
- codex --auto-edit
- codex --full-auto
Review outputs: Codex prints proposed patches and shell commands inline. Approve, reject, or tweak as desired.

FAQ

Which models does Codex use?

By default Codex targets o4‑mini for fast reasoning. You can specify any model available in the Responses API, e.g. codex -m o3. We’d also recommend verifying your developer account to start seeing chain of thought summaries in the API.

Does Codex upload my code to OpenAI?

No. All file reads, writes, and command executions happen locally. Only your prompt, high‑level context, and optional diff summaries are sent to the model for generation.

How do I change approval modes?

Start Codex with the appropriate flag (--suggest, --auto-edit, --full-auto) or use the in‑session shortcut /mode to toggle.

What platforms are supported?

Codex officially supports macOS and Linux. Windows support is experimental and may require WSL.

The agent is stuck or times out—what can I do?

Ensure you have a stable internet connection for API calls.
Try CTRL‑C to cancel the current step and ask Codex to continue.
If using Full Auto mode, confirm that the sandbox has permission to access required directories.