External Invocation

Integrate into a third-party application

Alibaba Cloud Model Studio supports configuring MCP services in Cherry Studio and Cursor. Both automatic and manual configuration methods are supported. The following example uses the Amap Maps service.

Develop integration using the SDK

You can use the MCP SDK to call Alibaba Cloud Model Studio MCP services. This provides full coding flexibility.

The following example uses the OpenAI SDK and MCP SDK to call the Model Studio WebSearch MCP service for web search.

1. Install the dependencies.

   pip install openai mcp

2. You can set your Alibaba Cloud Model Studio API key as an environment variable.

3. Sample code:

   Python

   # -*- coding: utf-8 -*-
   # Use OpenAI SDK + MCP SDK to call the Model Studio WebSearch MCP service
   import os
   import asyncio
   import json
   from openai import OpenAI
   from mcp.client.streamable_http import streamablehttp_client
   from mcp import ClientSession
   
   async def main():
       api_key = os.getenv("DASHSCOPE_API_KEY")
       if not api_key:
           print("Error: Set the DASHSCOPE_API_KEY environment variable")
           return
       mcp_url = "https://dashscope.aliyuncs.com/api/v1/mcps/WebSearch/mcp"
       headers = {"Authorization": f"Bearer {api_key}"}
       # 1. Connect to the MCP server and list available tools
       async with streamablehttp_client(mcp_url, headers=headers) as (read, write, _):
           async with ClientSession(read, write) as session:
               await session.initialize()
               tools_result = await session.list_tools()
               # Convert to OpenAI function calling format
               openai_tools = []
               for tool in tools_result.tools:
                   openai_tools.append({
                       "type": "function",
                       "function": {
                           "name": tool.name,
                           "description": tool.description or "",
                           "parameters": tool.inputSchema or {"type": "object", "properties": {}},
                       },
                   })
               # 2. Call DashScope (OpenAI-compatible endpoint)
               client = OpenAI(
                   api_key=api_key,
                   base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",
               )
               messages = [{"role": "user", "content": "Search for the latest updates on Alibaba Cloud Model Studio MCP"}]
               print("Searching the web...")
               print("=" * 50)
               # 3. Multi-turn tool call loop
               while True:
                   response = client.chat.completions.create(
                       model="qwen-max",
                       messages=messages,
                       tools=openai_tools or None,
                   )
                   choice = response.choices[0]
                   msg = choice.message
                   if not msg.tool_calls:
                       print(msg.content)
                       break
                   messages.append(msg)
                   for tc in msg.tool_calls:
                       args = json.loads(tc.function.arguments)
                       result = await session.call_tool(tc.function.name, args)
                       tool_content = ""
                       for block in result.content:
                           if hasattr(block, "text"):
                               tool_content += block.text
                       messages.append({
                           "role": "tool",
                           "tool_call_id": tc.id,
                           "content": tool_content,
                       })
   
   if __name__ == "__main__":
       asyncio.run(main())

4. You can run the code. The output appears as follows:

   Searching the web...
   ==================================================
   Alibaba Cloud Model Studio MCP (Model Context Protocol) is a newly launched service that allows users to uniformly access and manage MCP services on the Model Studio platform. Latest updates include:
   1. Supports one-click activation of various MCP services (such as Amap Maps, Web Search, etc.) through the MCP Marketplace.
   2. Provides the Streamable HTTP protocol, allowing external applications to call via standard HTTP.
   3. Integrated with mainstream tools such as Cherry Studio and Cursor, supporting automatic configuration.
   4. Developers can flexibly integrate MCP into their own projects via the MCP SDK.

What do I do if I cannot connect to an MCP service?

1. MCP service not enabled or not upgraded: Ensure the MCP service is enabled or upgraded in the Model Studio MCP Marketplace. See Enable an MCP service.

2. Invalid API key: Confirm that you are using a valid Alibaba Cloud Model Studio API key.

3. Quota exhausted: Some MCP services, such as web search, have monthly quotas. They stop automatically when the quota is exhausted.

For answers to common questions and information about other errors, see the FAQ.

The model responds normally and no MCP errors appear, but the MCP tool still does not run. What do I do?

The Large Language Model needs clear instructions to correctly call an MCP service. Specify the tool name and its capabilities in your prompt. For example, call the Alibaba Cloud Model Studio Amap Maps MCP service to plan a driving route from Hangzhou to Shanghai.