Qwen Code CLI

Qwen Code CLI can use OdiRouter as an OpenAI-compatible provider under the openai auth type. Declare the model in ~/.qwen/settings.json and provide the OdiRouter API key through an environment variable.

Requirements

Requirement	Value
Node.js	Required by Qwen Code CLI
OdiRouter API key	Create one in the OdiRouter console
Base URL	https://www.odirouter.ai/v1
Auth type	openai
Protocol	openai

Install Qwen Code CLI

macOS / Linux / WSL

npm install -g @qwen-code/qwen-code

Windows

npm install -g @qwen-code/qwen-code

Different Qwen Code versions may support different CLI flags. Prefer configuring the provider and model in ~/.qwen/settings.json; if you see context hard limit errors, first confirm the model entry includes generationConfig.contextWindowSize.

Create The Config File

macOS / Linux / WSL

mkdir -p ~/.qwen
nano ~/.qwen/settings.json

Windows

New-Item -ItemType Directory -Force "$env:USERPROFILE\.qwen"
notepad "$env:USERPROFILE\.qwen\settings.json"

Set The API Key

Do not hardcode the API key in settings.json. Qwen Code modelProviders entries can read credentials through envKey.

macOS / Linux / WSL

export ODIROUTER_API_KEY="YOUR_API_KEY"

Windows

$env:ODIROUTER_API_KEY = "YOUR_API_KEY"

Paste The OdiRouter Model

Replace the model ID with the OdiRouter model ID you want to expose in Qwen Code. Use a Qwen Code supported auth type as the modelProviders key, here openai; do not invent an odirouter auth type.

{
  "modelProviders": {
    "openai": {
      "protocol": "openai",
      "models": [
        {
          "id": "gpt-5.5",
          "name": "GPT 5.5 via OdiRouter",
          "baseUrl": "https://www.odirouter.ai/v1",
          "envKey": "ODIROUTER_API_KEY",
          "generationConfig": {
            "contextWindowSize": 128000,
            "samplingParams": {
              "max_tokens": 1024
            }
          }
        }
      ]
    }
  }
}

contextWindowSize tells Qwen Code the context window for this custom model. Without it, some Qwen Code versions may treat an unknown model’s hard limit as 0 and reject the request before sending it.

Start Qwen Code

cd /path/to/your/project
qwen

If Qwen Code asks you to choose a model, select the OdiRouter model from the configured provider. You can also run /model inside Qwen Code and choose GPT 5.5 via OdiRouter.

Verify The API Key

Before debugging Qwen Code, confirm that the key and model work outside the CLI:

curl -sS "https://www.odirouter.ai/v1/chat/completions" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.5",
    "messages": [
      {
        "role": "user",
        "content": "Reply with: Qwen Code connected to OdiRouter"
      }
    ]
  }'

Usage

Start with a read-only task:

Explain the structure of this repository and list the main entry points.

Then try a scoped edit:

Update the README with a short local development section. Show the diff before applying changes.

Troubleshooting

Symptom	Fix
401 or invalid key	Confirm ODIROUTER_API_KEY is available in the current shell
Model not found	Query GET https://www.odirouter.ai/v1/models and copy the exact model ID
Config ignored	Confirm the file path is ~/.qwen/settings.json on macOS/Linux/WSL or %USERPROFILE%\.qwen\settings.json on Windows
Model does not appear in /model	Confirm the modelProviders key is openai and the model is inside the models array
Context hard limit is 0	Set contextWindowSize in the model’s generationConfig
Streaming issues	Try another OdiRouter model or disable streaming if Qwen Code exposes that option