MCP Server Demo工程

What is MCP Server Demo工程?

MCP Server Demo工程 is a Spring AI-based MCP server that enables large language models to invoke Spark tasks via tools. It integrates with Alibaba Cloud Bailian's Qwen model through the Cline plugin in VS Code, providing a local server that accepts model requests, runs Spark jobs, and returns execution results. It is suitable for developers who want to experiment with MCP workflows without owning dedicated cloud server resources.

How to use MCP Server Demo工程?

Start the MCP server by running the run-sse-server.sh script. Configure Cline to use the Bailian Qwen model and point the MCP server configuration to the local server address. Once set up, ask the model a question like "Please use MCP Server to start a Spark task" in Cline. The model will request to call the MCP server; you can either approve each call or enable auto-approval. The server executes the Spark task and returns the result to the model, which then relays it to you.

Key features of MCP Server Demo工程

• Rapid MCP Server development using Spring AI
• Tool method for invoking and tracking Spark tasks
• Integration with Alibaba Cloud Bailian Qwen model via Cline
• Local SSE server with built-in client for testing
• Supports auto-approve or manual confirmation of tool calls

Use cases of MCP Server Demo工程

• Starting a Spark job by asking a language model in Cline
• Demonstrating an end-to-end MCP workflow with a real compute task
• Testing and prototyping MCP-based tool calling without cloud server overhead

FAQ from MCP Server Demo工程

How do I configure the MCP server in Cline?

Add a new MCP server entry in Cline's MCP configuration, set the transport to SSE, and point the URL to http://localhost:8080 (or the custom port). Then start the server using the provided run-sse-server.sh script.

What are the runtime requirements?

You need Java and the Spring AI framework. The server runs locally, and the client requires VS Code with the Cline extension and a valid Alibaba Cloud Bailian API token (trial or paid).

Is the data stored locally or sent to the cloud?

Data stays local. The MCP server runs on your machine, and the Spark task executes locally. Only model queries and tool call metadata are exchanged with the Bailian API.

Can I use a different model instead of Qwen?

The README specifically uses Bailian's Qwen model, but the architecture is model-agnostic. You can configure Cline to use any model provider that supports tool calling and SSE-based MCP servers.

Can I implement the MCP server without Spring AI?

Yes. The README notes that if you prefer not to use Spring AI, you can implement the same functionality directly using the MCP SDK.