Codex is OpenAI's official agent that helps you write, review and ship code. It runs as a CLI, an IDE extension, a desktop app and in the cloud, works on several tasks in parallel, and reads and edits files, runs commands and checks the results on your machine.
Already have an account? Sign in
CodexAbout 10 minutes: install Codex CLI → create a key in Portal → import it with CC-Switch (or edit config.toml) → start Codex.
Use the official standalone installer, npm, or Homebrew.
curl -fsSL https://chatgpt.com/codex/install.sh | shpowershell -ExecutionPolicy ByPass -c "irm https://chatgpt.com/codex/install.ps1 | iex"npm install -g @openai/codexbrew install --cask codexYou can also download a platform archive from GitHub Releases (e.g. codex-aarch64-apple-darwin.tar.gz), rename the extracted file to codex and put it on your PATH.
Open a new terminal and run the command below; it should print a version number:
codex --versionCodex providers live in ~/.codex/config.toml. CC-Switch can write it for you, or you can add it by hand.
sk-) and the API base URL. The key is shown in full only once, so keep it somewhere safe.CC-Switch is an open-source (MIT) desktop app that manages API providers for Claude Code, Codex, Gemini CLI, OpenCode and more. It writes the configuration files for you, so you don't have to edit them by hand. Requirements: Windows 10+, macOS 12+, or Linux x86_64/ARM64 with glibc 2.35+ and WebKitGTK 4.1 (e.g. Ubuntu 22.04+, Debian 12+).
Install with Homebrew (recommended):
brew install --cask cc-switchOr download CC-Switch-v{version}-macOS.dmg from Releases. It is a Universal build signed and notarized by Apple, so it runs on both Apple Silicon and Intel Macs.
Download CC-Switch-v{version}-Windows.msi (or the -Windows-Portable.zip portable build) from Releases and run it. On ARM Windows use the -Windows-arm64 files.
From Releases, download the .deb (Debian/Ubuntu), .rpm (Fedora etc.) or .AppImage for your architecture. For example on Ubuntu:
sudo apt install ./CC-Switch-v*-Linux-x86_64.debOn Arch Linux: paru -S cc-switch-bin.
ccswitch:// link handler on first launch).Prefer to add it by hand? In CC-Switch click the orange + → Custom, fill in Provider name (e.g. codelux.io), Website https://codelux.io, your API Key and the Request URL (the Codelux API base URL, no trailing slash), then click Add.
The Portal import sets the default model to gpt-5.5, and CC-Switch writes the Codelux provider (Responses protocol) to ~/.codex/config.toml. Close and reopen your terminal after switching providers for Codex to pick it up.
Open (or create) the user config file and add:
| OS | File |
|---|---|
| macOS / Linux | ~/.codex/config.toml |
| Windows | %USERPROFILE%\.codex\config.toml |
model = "gpt-5.5"
model_provider = "codelux"
[model_providers.codelux]
name = "Codelux"
base_url = "https://codelux.io/v1"
env_key = "CODELUX_API_KEY"
wire_api = "responses"model_provider points to the [model_providers.codelux] table below it.base_url is the Codelux API base URL with /v1; Codex appends /responses.env_key is the name of the environment variable that holds your key.wire_api = "responses" makes Codex use the Responses API, which Codelux supports.Then put the key in that environment variable (add it to your shell profile to keep it):
echo 'export CODELUX_API_KEY=sk-your-codelux-key' >> ~/.zshrc
source ~/.zshrcsetx CODELUX_API_KEY "sk-your-codelux-key"setx applies to terminals opened afterwards.
Start Codex in your project, confirm it uses the Codelux provider, then switch models as needed.
cd your-project
codexYou don't need Sign in with ChatGPT with a third-party provider. In the session, run /status to confirm the model and the Codelux provider, then send a test prompt.
Run /model in a session, or choose a model at launch:
codex --model gpt-5.5To change the default, edit model in config.toml. Model IDs are listed in Portal → AI Hub.
| Command | What it does |
|---|---|
/init | Create an AGENTS.md with project instructions |
/model | Switch model and reasoning effort |
/status | Show the current model, provider and token usage |
codex exec "..." | Run a task non-interactively (scripts and CI) |
See each request's model, tokens and charge in Portal → Logs.
Check that model_provider = "codelux" is at the top level of ~/.codex/config.toml (not inside a table) and that the environment variable named in env_key is set in the terminal you start Codex from.
The variable in env_key is empty in this terminal, or the key was deleted. Run echo $CODELUX_API_KEY to check, then reopen the terminal.
base_url must end with /v1 (e.g. https://codelux.io/v1), not /v1/responses.
Yes. Set model or run /model with any model ID from Portal → AI Hub; Codelux handles the Responses API for it.