Qwen Code

What is Qwen Code

Qwen Code is an open-source coding agent for the terminal. Its modelProviders setting lists the models you can use and the API protocol of each. This setup adds Capriole AI models to that list.

Prerequisites

Before you begin, make sure you have:

A Capriole AI account (Sign up here)
A Capriole AI API key (Get your key)

Installation

Install Qwen Code with npm (Node.js 22 or later):

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

Or install it with Homebrew:

brew install qwen-code

Verify the install:

qwen --version

Configuration

Model validation

Qwen Code 0.25.3 passes a file-read, file-write, and final-answer check through Capriole AI with every model in Supported models, using the configuration below. Put your key in ~/.qwen/.env:

CAPRIOLE_AI_API_KEY=YOUR_CAPRIOLE_AI_API_KEY

Create or update ~/.qwen/settings.json:

{
  "security": {
    "auth": {
      "selectedType": "openai"
    }
  },
  "model": {
    "name": "openai-latest"
  },
  "modelProviders": {
    "openai": [
      {
        "id": "openai-latest",
        "name": "openai-latest",
        "baseUrl": "https://api.caprioletech.com/v1",
        "envKey": "CAPRIOLE_AI_API_KEY",
        "generationConfig": {
          "contextWindowSize": 1050000
        },
        "wireApi": "responses"
      },
      {
        "id": "openai/gpt-6.1-sol",
        "name": "openai/gpt-6.1-sol",
        "baseUrl": "https://api.caprioletech.com/v1",
        "envKey": "CAPRIOLE_AI_API_KEY",
        "generationConfig": {
          "contextWindowSize": 1050000
        },
        "wireApi": "responses"
      },
      {
        "id": "zai/glm-5.3-flash",
        "name": "zai/glm-5.3-flash",
        "baseUrl": "https://api.caprioletech.com/v1",
        "envKey": "CAPRIOLE_AI_API_KEY",
        "generationConfig": {
          "contextWindowSize": 1000000
        }
      },
      {
        "id": "moonshot/kimi-k3",
        "name": "moonshot/kimi-k3",
        "baseUrl": "https://api.caprioletech.com/v1",
        "envKey": "CAPRIOLE_AI_API_KEY",
        "generationConfig": {
          "contextWindowSize": 1048576
        }
      },
      {
        "id": "xai/grok-4.7",
        "name": "xai/grok-4.7",
        "baseUrl": "https://api.caprioletech.com/v1",
        "envKey": "CAPRIOLE_AI_API_KEY",
        "generationConfig": {
          "contextWindowSize": 500000
        }
      },
      {
        "id": "google/gemini-3.8-flash",
        "name": "google/gemini-3.8-flash",
        "baseUrl": "https://api.caprioletech.com/v1",
        "envKey": "CAPRIOLE_AI_API_KEY",
        "generationConfig": {
          "contextWindowSize": 1048576
        }
      }
    ],
    "anthropic": [
      {
        "id": "claude-latest",
        "name": "claude-latest",
        "baseUrl": "https://api.caprioletech.com",
        "envKey": "CAPRIOLE_AI_API_KEY",
        "generationConfig": {
          "contextWindowSize": 1000000
        }
      },
      {
        "id": "claude-opus-5-5",
        "name": "claude-opus-5-5",
        "baseUrl": "https://api.caprioletech.com",
        "envKey": "CAPRIOLE_AI_API_KEY",
        "generationConfig": {
          "contextWindowSize": 1000000
        }
      }
    ]
  }
}

What each setting does:

Models under openai use Chat Completions. GPT models add wireApi: "responses" to use the Responses API.
Models under anthropic use Anthropic Messages. Their baseUrl has no /v1.
envKey names the environment variable that holds your key. Qwen Code sends the key as a Bearer token to Capriole AI.
contextWindowSize tells Qwen Code each model’s context window.
security.auth.selectedType and model.name set the default model. /model updates both when you switch.

Run Qwen Code

Start Qwen Code from your project directory:

qwen

Switch models with /model. For a one-shot run, pass the prompt, the model, and the auth type of its group:

qwen "Explain this repository." -m claude-latest --auth-type anthropic

Supported models

Model	Model ID	Group	Protocol
OpenAI latest	`openai-latest`	`openai`	Responses
GPT-6.1 Sol	`openai/gpt-6.1-sol`	`openai`	Responses
Claude latest	`claude-latest`	`anthropic`	Anthropic Messages
Opus 5.5	`claude-opus-5-5`	`anthropic`	Anthropic Messages
Gemini 3.8 Flash	`google/gemini-3.8-flash`	`openai`	Chat Completions
Grok 4.7	`xai/grok-4.7`	`openai`	Chat Completions
GLM 5.3 Flash	`zai/glm-5.3-flash`	`openai`	Chat Completions
Kimi K3	`moonshot/kimi-k3`	`openai`	Chat Completions

openai-latest and claude-latest always point to our latest flagship in each family.

Troubleshooting

Why does Qwen Code return an authentication error?

Confirm that ~/.qwen/.env sets CAPRIOLE_AI_API_KEY, or export the variable in the same shell.

Why does a model fail with a protocol error?

Run the model with the auth type of its group: --auth-type openai for models under openai, and --auth-type anthropic for models under anthropic. /model sets this for you.

Why do Claude requests fail with `Unsupported Claude Messages model`?

Claude models under anthropic need the bare Claude IDs, such as claude-opus-5-5.

Notes

Keep your key in ~/.qwen/.env, not in settings.json.
To add another Capriole AI model, copy the entry of a model from the same family and change the ID. GET https://api.caprioletech.com/v1/models lists every model ID.
Official references: Qwen Code and Qwen Code model providers.

Get started

Endpoints

Integration

What is Qwen Code

Prerequisites

Installation

Configuration

Model validation

Run Qwen Code

Supported models

Troubleshooting

Why does Qwen Code return an authentication error?

Why does a model fail with a protocol error?

Why do Claude requests fail with `Unsupported Claude Messages model`?

Notes

​What is Qwen Code

​Prerequisites

​Installation

​Configuration

​Model validation

​Run Qwen Code

​Supported models

​Troubleshooting

​Why does Qwen Code return an authentication error?

​Why does a model fail with a protocol error?

​Why do Claude requests fail with Unsupported Claude Messages model?

​Notes

What is Qwen Code

Prerequisites

Installation

Configuration

Model validation

Run Qwen Code

Supported models

Troubleshooting

Why does Qwen Code return an authentication error?

Why does a model fail with a protocol error?

Why do Claude requests fail with `Unsupported Claude Messages model`?

Notes