Qwen Code + Token Harbor

Export TOKENHARBOR_API_KEY=YOUR_thk_KEY, then merge this into ~/.qwen/settings.json:

{
  "modelProviders": {
    "openai": [{
      "id": "deepseek-v4.1-flash:free",
      "name": "DeepSeek V4.1 Flash (Free)",
      "baseUrl": "https://tokenharbor.ai/v1",
      "envKey": "TOKENHARBOR_API_KEY"
    }]
  },
  "security": { "auth": { "selectedType": "openai" } },
  "model": { "name": "deepseek-v4.1-flash:free" }
}

The current provider value is an array. The preview wrapper { "protocol": "openai", "models": [...] } is ignored in migrated settings; Connect 0.6.4 converts that wrapper and preserves a backup. Keys stay in the environment. Open a new terminal, run qwen, and use /model to switch.

Read the official model-provider guide. Use an ID returned by your authenticated /v1/models request for other models.