Codex CLI Manual Configuration

Make sure you have completed the environment check before continuing with the steps below.

1. Open the configuration directory

%userprofile%\.codex

open ~/.codex        # macOS
xdg-open ~/.codex    # Linux

Codex uses three main files:

If these files don't exist in the directory, create them manually.

2. Write config.toml

disable_response_storage = true
model = "gpt-5.3-codex"
model_provider = "getmodel"
model_reasoning_effort = "xhigh"
model_verbosity = "high"

[features]
web_search_request = true

[model_providers.getmodel]
name = "getmodel"
base_url = "https://getmodel.ai/v1"
requires_openai_auth = true
wire_api = "responses"

Key points:

• The model_provider name (here getmodel) must match the [model_providers.xxx] block name
• The base_url should follow the actual address shown in the Console
• Change model to the specific model name you want to use by default

3. Write auth.json

{
  "OPENAI_API_KEY": "你的-API-令牌"
}

Replace "你的-API-令牌" with the string you obtained in Create a token.

4. Verify

Run in the terminal:

codex

Once you're in the chat interface, send a message to test. If you get a normal reply, the configuration was successful.