Codex Install, Configure, and Uninstall Guide: CLI, App, and IDE Extension Tutorial for 2026

The real point: Codex is not finished after installation

When people read a Codex tutorial for the first time, they usually focus on “how to install it.” But once you start using it, installation is only the first step.

Codex behaves more like an AI coding agent that works inside your project directory. It can read code, edit files, run commands, and affect a real repository. So the right setup is not “install and go.” You need accounts, permissions, configuration, Git checkpoints, and clear boundaries first.

In other words, Codex is not just an installer. It is a development workflow: install → sign in → configure → enter a project → run a task → review the result → merge or roll back.

Choose the right entry point: CLI, App, IDE extension, or Cloud

Codex is no longer a single interface. You can use Codex CLI in the terminal, manage work in the macOS / Windows Codex App, use the IDE extension in editors such as VS Code, Cursor, and Windsurf, or delegate work through Codex in the cloud.

The choice is simple. If you live in the terminal, use Codex CLI. If you want multi-threaded project management, visual review, and project switching, use the Codex App. If you spend all day inside your editor, the IDE extension will feel more natural.

That is the useful structure of the original article: do not only explain one command. Explain installation, configuration, uninstall, and troubleshooting as one setup flow.

What to prepare before installing

Before installation, check four things: account, network, project, and rollback. Missing one of them can turn into confusing errors later.

For the account, prepare a ChatGPT / OpenAI account that can access Codex, depending on your plan and organization settings. For the project, use Codex inside a Git repository whenever possible, and keep checkpoints before and after important tasks.

For the network, keep it stable instead of switching locations during sign-in. For configuration, understand the user-level and project-level config locations, especially ~/.codex/config.toml and .codex/config.toml inside a repository.

How to install Codex CLI

If you want a terminal-first workflow, Codex CLI is the most direct entry point. The official docs offer several installation paths: standalone installer, npm, and Homebrew.

On macOS or Linux, you can use the official shell installer. On Windows, you can use the PowerShell installer. If you already rely on Node.js, npm works well. If you are a macOS developer, Homebrew is often the cleanest option.

After installation, do not immediately let Codex rewrite a real project. Open a test directory, run codex, sign in, and ask it to explain the folder or create a small file. Confirm that path, permissions, and authentication all work first.

Quick command reference

macOS / Linux standalone installer

curl -fsSL https://chatgpt.com/codex/install.sh | sh

Windows PowerShell installer

powershell -ExecutionPolicy ByPass -c "irm https://chatgpt.com/codex/install.ps1 | iex"

npm install

npm install -g @openai/codex

Homebrew install

brew install --cask codex

Run

codex

npm uninstall

npm uninstall -g @openai/codex

Homebrew uninstall

brew uninstall --cask codex

Carefully remove local configuration

rm -rf ~/.codex

How to choose between Codex App and the IDE extension

Codex App is better when you want to manage multiple tasks in one place. It fits parallel threads, review, worktrees, browser preview, and project switching.

The IDE extension is better for “ask while editing.” It keeps you inside the editor, which is useful for fixing bugs, adding tests, explaining functions, and refactoring small modules.

My suggestion is simple: developers can start with CLI or the IDE extension. Team leads and people who review several projects can add the Codex App later. Do not install every interface on day one if it makes the workflow messy.

Configuration: config.toml, permissions, and MCP

Codex configuration should not live in random chat messages. OpenAI’s docs state that user-level configuration is stored at ~/.codex/config.toml, and repository-specific rules can live in .codex/config.toml inside the project.

This matters because an AI coding agent does more than answer questions. It can read files, run commands, and write code. Which folders it can read, which commands require confirmation, and which tools it can call should be explicit.

If you connect MCP servers, treat that as real configuration management too. MCP can connect Codex to more tools, but more tools also mean stronger permission boundaries and better review loops.

Common issues: the environment is usually the problem

Most beginner issues are not caused by Codex being broken. They come from inconsistent environments: PATH not refreshed, both npm and Homebrew versions installed, Windows permissions, WSL and native Windows paths mixed together, or terminal authentication not matching the app session.

The debugging order is simple: check codex --version, then check which codex or where codex, and confirm whether your active binary came from npm, brew, standalone installer, or the app workflow.

On Windows, pay special attention to sandbox mode, winget, administrator approval, and WSL2. Use the native Windows path first when possible, and use WSL2 when your development workflow really needs Linux tooling.

Uninstalling: remove the app, or remove the configuration too?

When uninstalling Codex, do not delete every folder first. Decide whether you only want to remove the program, or whether you also want to remove authentication, configuration, and local history.

If you installed with npm, use npm uninstall -g @openai/codex. If you installed with Homebrew, use brew uninstall --cask codex. If you used a standalone installer, remove the binary from the actual install path. For Codex App, uninstall it like a normal system app.

Only delete ~/.codex if you intentionally want to remove local settings. That can clear configuration, MCP setup, login state, and local preferences, so do it carefully.

How this connects to We0.ai

Codex helps turn intent into code. But code is not the finish line. To reach customers, you still need to show product value, cases, documentation, tutorials, and trust.

That is where We0.ai fits. We0.ai is not just a normal AI website builder. It is an AI Showcase Website Growth Platform. It focuses on the Build → Showcase → Grow → Leads path: build the product, explain it clearly, get discovered by search and AI answers, and convert visitors into leads.

A stronger workflow is this: use Codex to improve development and delivery, then use We0.ai to turn product features, services, cases, and tutorials into a searchable, understandable, conversion-ready showcase website.

Final takeaway

If you only want to try Codex, following the install commands is enough.

But if you want to put it into a real development process, do not focus only on installation. Design a stable workflow: fixed entry point, clear configuration, Git rollback, permission boundaries, and regular cleanup.

AI coding tools will keep getting stronger. But the most valuable teams will not be the ones with the most tools. They will be the ones that connect tools to real business workflows. Code should run. Products should be shown. Content should be discoverable. Websites should create leads.

FAQ

What is the difference between Codex CLI and Codex App?

Codex CLI is better for terminal-first workflows where Codex reads, edits, and runs code in the current project directory. Codex App is better for visual task management, review, worktrees, and project switching.

Do I have to install Codex with npm?

No. The official docs support standalone installers, npm, and Homebrew. Choose one primary method that matches your development environment, and avoid mixing several installation methods.

Where is the Codex configuration file?

User-level configuration is usually stored at ~/.codex/config.toml. Repository-specific configuration can live in .codex/config.toml inside a project. The CLI and IDE extension share these configuration layers.

Should Windows users choose native Windows or WSL2?

Use the native Windows workflow when it works. Use WSL2 when your repository and tooling already depend on a Linux-native environment.

Should I delete ~/.codex when uninstalling Codex?

Only delete it when you intentionally want to remove local configuration, MCP setup, login state, and preferences. A normal uninstall does not always require deleting this folder.

Why does a Codex guide mention We0.ai?

Codex helps with code delivery. We0.ai helps with showcase and growth. A complete product workflow is not only about writing code; it also turns product pages, cases, docs, and tutorials into a website that search engines and AI answers can understand.

Related Tools

- OpenAI Codex

- Codex CLI

- Codex App

- Codex IDE Extension

- GitHub

- Visual Studio Code

- Homebrew

- Node.js

- We0.ai

Sources

- Original CSDN Article

- OpenAI Codex

- Codex CLI

- Codex Quickstart

- Codex App

- Codex Configuration Basics

- Codex Configuration Reference

- Codex Windows Guide

- OpenAI Codex GitHub Repository