Skip to content

Codex Setup

Codex is OpenAI's command-line coding tool. To use it through Yunxi AI, configure a Yunxi AI model provider and authenticate with a dedicated Codex key.

Prerequisites

  1. You have signed in to https://ai.laiber.cloud.
  2. You have created a Codex API key.
  3. Node.js and npm are installed.

Recommended

If CC-Switch is installed, open the key action menu, choose CC Switch, select Codex, and import the model profile directly. The config.toml method below is a manual fallback.

Install Codex

bash
npm install -g @openai/codex@latest

Verify:

bash
codex --version

Create the config directory

bash
mkdir -p ~/.codex
cd ~/.codex

Write config.toml

The example below uses the OpenAI Responses-compatible endpoint. Adjust model based on models available to your account.

toml
model_provider = "yunxi"
model = "gpt-5.5"
model_reasoning_effort = "high"
disable_response_storage = true

[model_providers.yunxi]
name = "Yunxi AI"
base_url = "https://ai.laiber.cloud/v1"
wire_api = "responses"
requires_openai_auth = true

Write auth.json

Replace sk-xxx with your Codex key:

json
{
  "OPENAI_API_KEY": "sk-xxx"
}

Start Codex

bash
cd your-project
codex

Ask Codex to inspect the current directory or create a small test file to confirm requests are working.

FAQ

Can Codex and Claude Code share one key?

Not recommended. Their protocols and model groups may differ, so separate keys are easier to debug and revoke.

Where are config files stored?

By default, Codex reads from ~/.codex.

Stable, clear, and easy to connect