Qwen Code setup

Qwen Code's openai protocol supports OpenAI-compatible endpoints. These launch arguments provide the authentication method, model, and credential directly, without entering a sign-in flow or writing a settings file:

qwen --auth-type openai \
  --model "your-client-facing-model-name" \
  --openai-api-key "your-lumoswitch-access-key" \
  --openai-base-url "https://api.lumoswitch.com/v1"

For multiple persistent models, use modelProviders in user-level ~/.qwen/settings.json.

Export the credential in the shell that starts Qwen Code:

export OPENAI_API_KEY="your-lumoswitch-access-key"

{
  "modelProviders": {
    "openai": {
      "protocol": "openai",
      "models": [
        {
          "id": "your-client-facing-model-name",
          "name": "Lumoswitch model",
          "envKey": "OPENAI_API_KEY",
          "baseUrl": "https://api.lumoswitch.com/v1"
        }
      ]
    }
  }
}

Project-level .qwen/settings.json overrides user settings. Check this precedence and the current shell environment if setup appears ignored.

References: Qwen Code authentication and model providers.

Qwen Code setup

On this page