Qwen Chat Integration

Claude Code Router supports routing through Qwen Chat (通义千问) via the qwen.aikit.club API. It uses JWT-based authentication where you paste a token copied from the chat.qwen.ai web application.

How It Works

1. The CCR server hosts an auth page at http://localhost:3456/qwen/auth
2. The page offers a bookmarklet and a manual paste form for obtaining the JWT token
3. Once submitted, the token is validated against Qwen's API and saved to ~/.claude-code-router/qwen_auth.json
4. Automatic token rotation — expired tokens are detected on the next request and you'll be prompted to refresh
5. Qwen's trailing <details>...</details> metadata block is stripped from responses automatically

Prerequisites

• Access to the Qwen Chat API via qwen.aikit.club
• A Qwen Chat account

Setup

1. Authenticate via Browser

The CCR server hosts an auth page at http://localhost:3456/qwen/auth. Open it in your browser:

http://localhost:3456/qwen/auth

The page offers two methods:

Option 1 — Bookmarklet (recommended):

1. On the auth page, drag the "Get Qwen Token" button to your bookmarks bar
2. Open chat.qwen.ai in another tab and sign in
3. Click the bookmarklet on the Qwen page — it reads the token from localStorage and sends it back to the auth page automatically

Option 2 — Manual paste:

1. Open chat.qwen.ai, sign in, and open DevTools (F12)
2. In the Console tab, run: copy(localStorage.getItem('token'))
3. Go back to the auth page and paste the token into the form

The token is validated against Qwen's API and saved to ~/.claude-code-router/qwen-auth.json.

3. Configure Provider

Add the Qwen provider to your ~/.claude-code-router/config.json:

{
  "Providers": [
    {
      "name": "qwen",
      "baseUrl": "https://qwen.aikit.club/v1/chat/completions",
      "apiKey": "oauth_dummy_key",
      "models": ["qwen-max", "qwen-plus", "qwen-turbo"],
      "transformer": {
        "use": ["qwen-auth", "reasoning", "OpenAI"]
      }
    }
  ],
  "Router": {
    "default": "qwen,qwen-max"
  }
}

The qwen-auth transformer handles:

• Adding the Authorization: Bearer header with your token
• Stripping the <details>...</details> metadata block from responses

The reasoning transformer captures and replays Qwen's reasoning content across multi-turn conversations, maintaining coherence.

The OpenAI transformer registers the /v1/chat/completions endpoint — Qwen uses standard Chat Completions format.

4. Restart

docker compose restart ccr

Model Reference

Model	Description
qwen-max	Flagship model, best quality
qwen-plus	Balanced performance and cost
qwen-turbo	Fast, cost-effective for simple tasks

Troubleshooting

Invalid token: The JWT token may have expired. Re-run ccr qwen-auth and paste a fresh token from chat.qwen.ai.

Trailing metadata in responses: The <details>...</details> block should be stripped automatically. If you see it in raw responses, the qwen-auth transformer may not be active — check that it's in your provider's transformer.use array.