๐ Preparatory Steps
Before you begin, please ensure you have:
- Registered a Koze account and logged in
- Created an API Key on the API Management page
- Installed the Codex CLI (
npm install -g @openai/codex)
๐ Step 1: Configure the Codex Configuration File
Add the following configuration to the beginning of the ~/.codex/config.toml file:
model_provider = "koze"
model = "gpt-4.1"
model_reasoning_effort = "high"
disable_response_storage = true
preferred_auth_method = "apikey"
[model_providers.koze]
name = "koze"
base_url = "https://api.kozeai.com/v1"
wire_api = "responses"
requires_openai_auth = true
๐ก Tip: Replace `
https://api.kozeai.com/v1` with your actual Koze proxy address. The `model` field can select different models according to your plan.
๐ Step Two: Configure the API Key
Configure your API key in the ~/.codex/auth.json file:
{
'OPENAI_API_KEY': 'Your Koze API Key'
}
๐ฏ Step Three: Start Codex
After configuration, start it directly in the terminal:
codexCodex CLI will automatically use the Koze proxy for API requests.
๐ Available Models
You can use the following models in the `model` field of the configuration file (depending on your Koze plan):
| Model Name | Description |
|---|---|
gpt-5.3-codex |
gpt-5.3-codex, suitable for everyday programming tasks |
gpt-5.4-mini |
GPT-4.1 Mini, better value for money |
gpt-5.4 |
GPT-5.4 Multimodal Model |
โ ๏ธ Frequently Asked Questions
| Problem | Solution |
|---|---|
| Tip: "Authentication failed" | Check if the API Key in auth.json is correct |
config.toml is not working |
Ensure that the configuration is written at the beginning of the file and that the model_provider name matches [model_providers.xxx] |
| Using Nginx Anti-prototype fails. | Add `underscores_in_headers on;` to the Nginx ` |
ย