Unlock AI Data Integration with Qwen3, MCP & ComfyUI for Automated Content Creation

This article introduces the latest open‑source Qwen3‑235B‑A22B model and examines how Model Context Protocol (MCP) can be used to connect large models with various data sources and services, enabling end‑to‑end AI workflows such as image generation, storage, and automatic publishing to social platforms.

Architecture Overview

The MCP architecture allows multiple MCP servers to be orchestrated by a large model, providing high scalability and ease of use compared with multi‑agent solutions.

Practical Scheme

Scenario Introduction

MCP Server

FileSystem (available)

Xiaohongshu (personal deployment)

ComfyUI (personal deployment)

Large Model: Qwen3‑235B‑A22B

Prompt Content Examples

一只小狗在草地上奔跑

一群小朋友在欢乐的玩耍

写实风格的山水画

Effect Demonstration

Generated images are saved locally, uploaded to cloud storage, and automatically posted to platforms such as Xiaohongshu, Douyin, and WeChat.

Environment Preparation

Python 3.10 or higher

MCP – follow the official quick‑start guide

ComfyUI – can be deployed locally or on the cloud (see GitHub repo)

Qwen3 Integration

Install the CLine plugin in VSCode

Configure CLine to use the OpenAI‑compatible endpoint:

Base URL: https://dashscope.aliyuncs.com/compatible-mode/v1
Model ID: qwen3-235b-a22b

Server Code (FastMCP)

import urllib
import uuid
from typing import Dict, Any, List
import httpx
import logging
import websockets
from PIL import Image
import io
from fastmcp import FastMCP
from fastmcp.prompts import UserMessage
import json
from time import sleep

logging.basicConfig(level=logging.INFO)
logger = logging.getLogger("comfy-image-mcp-server")
COMFY_SERVER = "8.147.113.150:8000"
CLIENT_ID = str(uuid.uuid4())

mcp = FastMCP(name="CompyImageServer")

@mcp.prompt()
def generate_image_request(prompt: str, style: str = "动漫风格") -> UserMessage:
    content = f"生成一个comfyui的英文prompt，要求包含下面的内容: {prompt} 并且要求生成的图片需要具有很高的质量，风格是{style}"
    return UserMessage(content=content)

@mcp.tool()
def generate_image_async(prompt: str = "a cat with yellow hat", width=512, height=512, seed=4787458) -> Dict[str, Any]:
    workflow = { ... }  # omitted for brevity
    return queue_prompt(workflow)

if __name__ == "__main__":
    mcp.run(transport="sse", host="127.0.0.1", port=9000)

Launch Command

fastmcp run server.py:mcp --transport sse --host 127.0.0.1 --port 9000

Third‑Party Media Server

Uses the social‑auto‑upload project to publish generated content to Douyin, Bilibili, Xiaohongshu, etc.

CLINE Configuration (JSON)

{
  "mcpServers": {
    "filesystem": {
      "autoApprove": ["read_file", "list_allowed_directories", "read_multiple_files", "create_directory", "list_directory", "directory_tree", "search_files", "get_file_info"],
      "disabled": false,
      "timeout": 60,
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/Users/wangrupeng/Documents/Cline/MCP/filesystem-server"],
      "transportType": "stdio"
    },
    "comfyui": {
      "autoApprove": ["generate_image_async", "get_image_status", "get_image_status_and_download_to_local"],
      "disabled": false,
      "timeout": 60,
      "url": "http://127.0.0.1:9001/sse",
      "transportType": "sse"
    },
    "xiaohongshu": {
      "disabled": false,
      "timeout": 60,
      "url": "http://127.0.0.1:9002/sse",
      "transportType": "sse",
      "autoApprove": ["publish_xiaohongshu_note"]
    }
  }
}

Reflections and Thoughts

The core of both big data and AI is data. Any system with I/O—whether biological or computational—can be abstracted into a model where MCP bridges the context between humans, large models, and the physical world.

Potential commercial scenarios include selling MCP Server APIs, automating order analysis, enabling assistive robotics, and empowering embodied AI agents. However, challenges remain: large models are not yet fully reliable, MCP development can be complex, and security risks such as unsafe command execution and credential leakage must be addressed.

Current Development Tips

Prefer the FastMCP framework for concise server implementation.

Give tools clear, descriptive names and include prompt annotations.

Provide default values for all parameters to avoid missing arguments.

Future Outlook

Continued improvement of large‑model intelligence (e.g., newer Qwen releases).

Proliferation of MCP‑based applications and servers.

Urgent need to resolve security issues for reliable commercial deployment.

References

Bailian Console – AccessKey registration

MCP Filesystem Server documentation