⚠️ DISCLAIMER - This might be against Qwen Studio ToS, Use at your own risk ⚠️

Qwen-Scrape

A local OpenAI-compatible API bridge that allows AI coding agents (Cline, Roo Code, Continue.dev, Cursor, and custom Python agents) to use Qwen Studio models through a local network endpoint.

This project exposes an OpenAI-style API (/v1/chat/completions, /v1/models) and forwards requests to Qwen Studio by controlling an authenticated browser session through lightweight DOM and network automation.

Features

• ✅ Fully OpenAI-compatible API endpoint
• ✅ Works seamlessly with coding agents supporting custom OpenAI providers
• ✅ Supports dynamic Qwen Studio web model switching (e.g., Qwen3.7-Max, Qwen3.7-Plus)
• ✅ Real-time streaming response support
• ✅ Localhost (accessible from 127.0.0.1)
• ✅ No local model hosting or GPU required
• ✅ Compatible with OpenAI SDK / LangChain style clients

Architecture

Coding Agent (VS Code, Python, etc.)
     |
     | OpenAI API format (HTTP)
     v
Bridge Server (http://127.0.0.1:8000/v1)
     |
     | Command Queue & Stream Relay
     v
Browser Automation Script (Console Injection)
     |
     | Native DOM Injection & Network Interception
     v
Qwen Studio Web Interface (Authenticated Session)

Requirements

• Python 3.10+
• A valid, logged-in Qwen Studio session in a Chromium-based browser (Chrome, Edge, Brave)
• Tampermonkey extension (Recommended) or manual Console access

Installation

1. Clone the repository:

   git clone https://github.com/Vxtzq/qwen-scrape.git
   cd qwen-openai-bridge

2. Install Python dependencies:

   pip install -r requirements.txt

Usage

1. Start the Bridge Server

python server.py

You should see:

🟢 BRIDGE ACTIVE.
🔑 API Key: sk-qwen-bridge-key

2. Connect the Browser Session

Paste the content inside paste_in_chrome.js in the browser's console. That's all.

3. Test everything works

python benchmark_api.py

API Configuration

The bridge exposes its API on your local network. Find your local IP (e.g., 192.168.1.38) and configure your agents as follows:

Setting	Value
Base URL	http://127.0.0.1:8000/v1
API Key	sk-qwen-bridge-key
Model	qwen/qwen3.7-plus or qwen/qwen3.8-max

Supported Clients

Claude code / Qwen code

1. Type /auth
2. Select "Custom provider"
3. Select "OpenAI-Compatible"
4. Enter http://127.0.0.1:8000/v1
5. Enter sk-qwen-bridge-key
6. Add models (qwen/qwen3.7-plus or/and qwen/qwen3.8-max)
7. Press enter (you can enable thinking...)
8. Enjoy!

Cline / Roo Code (VS Code)

1. Open Cline Settings → API Provider.
2. Select OpenAI Compatible.
3. Enter the Base URL, API Key, and Model ID from the table above.
4. Click Verify Connection.

Continue.dev

Add to your config.json:

{
  "models": [
    {
      "title": "Qwen Bridge (Max)",
      "provider": "openai",
      "model": "qwen/qwen3.8-max",
      "apiBase": "http://192.168.1.38:8000/v1",
      "apiKey": "sk-qwen-bridge-key"
    }
  ]
}

Cursor

1. Go to Cursor Settings → Models.
2. Click Add OpenAI Compatible Model.
3. Fill in the Name, Base URL, API Key, and Model Name.

Python Example

from openai import OpenAI

client = OpenAI(
    base_url="http://192.168.1.38:8000/v1",
    api_key="sk-qwen-bridge-key"
)

response = client.chat.completions.create(
    model="qwen/qwen3.7-max",
    messages=[{"role": "user", "content": "Write a Python script to parse JSON."}],
    stream=True
)

for chunk in response:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)

Troubleshooting

Connection Failed

• Ensure server.py is running.
• Ensure the Qwen browser tab is open and the Tampermonkey script is active.
• Verify your Base URL includes /v1 (e.g., http://192.168.1.38:8000/v1).

Model Not Found

• Test the discovery endpoint in your browser: http://192.168.1.38:8000/v1/models
• Ensure the model string in your agent exactly matches one of the returned id values (e.g., qwen/qwen3.7-max).

Limitations

• Based on a web interface, might not be as reliable as a standard paid API
• Reasonning content is not available to this day
• Might be slower than a standard LLM API

Security Notes

⚠️ This project is intended for local, trusted network use only. Do not expose port 8000 to the public internet. Your browser session contains authenticated access to your Qwen account.

Disclaimer

This project interacts with a third-party web interface through browser automation and network interception. Use responsibly and respect the Terms of Service of the platforms you connect to. This is an unofficial community tool.

License

MIT License