OpenAI Codex

Codex is OpenAI’s coding agent that helps you write, review, and ship code faster. Use it side-by-side in your IDE or delegate larger tasks to the cloud.

You can view Help Guide for more details.

Install Codex

You can install Codex using one of the following methods.

Option 1: Install via Homebrew

brew install codex

Option 2: Install via npm

npm install -g @openai/codex

Configure Codex

Codex reads its settings from a local configuration file written in TOML format.

Step 1: Open the configuration file

vim ~/.codex/config.toml

Step 2: Define the default model and provider

# Default model used by Codex
model = "o3"

# Default model provider
model_provider = "openai-chat-completions"

Step 3: Configure the model provider

[model_providers.openai-chat-completions]
# Display name shown in the Codex interface
name = "gateway.theturbo"

# Base URL of the OpenAI-compatible API
base_url = "https://gateway.theturbo.ai"

# Environment variable that stores the API key
env_key = "YOUR_AI_GATEWAY_API_KEY"

# API protocol type (optional, defaults to "chat")
wire_api = "chat"

Set the API Key

Set your API key as an environment variable:

export YOUR_AI_GATEWAY_API_KEY="sk-xxxxxxxxxxxxxx"

This key is generated when you create an AI Gateway instance.

Use Codex

After configuration, you can start using Codex immediately.

Start an interactive session

codex

Send a one-off request

codex "Summarize the purpose of this repository."

All requests will be forwarded through the AI Gateway, allowing centralized authentication, routing, and usage tracking.

Notes

• Codex requires an OpenAI-compatible API. AI Gateway provides this interface.

• API keys should always be stored as environment variables, not hard-coded in config files.

• You may change the default model or provider at any time without reinstalling Codex.