Simple MCP Server C51

🚀 Simple MCP Server

This is an example project that demonstrates how to build an MCP server in a minimalist way.

✨ Features

• Weather Query: Utilize the Weather Query API provided by the Gaode Open Platform to implement MCP server functions, supporting the retrieval of real-time weather and future weather forecasts.
• DingTalk Group Messages: Send information to a specified DingTalk group via the DingTalk group robot.

🎥 Feature Demonstration

When a user asks, "Do I need to wear a cotton-padded jacket in Xi'an today?"

1. The AI automatically determines whether to call the MCP server.
2. The AI provides the latitude and longitude information of Xi'an according to the requirements of the MCP server.
3. The MCP server internally uses the Gaode Weather Query API to obtain weather data and returns the results to the AI.
4. After organizing the weather information, the AI sends it to the user.

Meanwhile, the AI can also send the weather information to the DingTalk group via the DingTalk group robot. The effect is as follows:

🛠️ Environment Configuration

Building and Running the MCP Server

• Node.js 22+: Ensure that the latest version of Node.js is installed.
• Pnpm: Use Pnpm for dependency management.

MCP Client and Large Model

It is recommended to use an AI dialogue client that supports the MCP protocol, such as Cherry Studio. In this example, the Qwen2.5-7B-Instruct model provided by Guiji Liudong is used. You need to register on Guiji Liudong, obtain an API key, and configure it in Cherry Studio.

⚠️ Important Note

Choose a large model with the Function Call feature. These models are usually marked (e.g., 🔧).

🔨 Building Steps

1. Clone the project code to a local path, such as /Users/yourname/simple-mcp-server/.
2. Open the terminal and switch to this directory.
3. Install dependencies: Run the command pnpm i.
4. Build the server: Run the command pnpm build.

After the build is complete, the build/index.js file will be generated in the project directory.

🚀 Server Installation and Configuration

Add the MCP server in Cherry Studio:

1. Open Settings > MCP Server.
2. Add a new server:
   • Name: Customize the name.
   • Type: Select STDIO.
   • Command: Specify the Node.js execution path.
   • Parameters: Enter the path of the build/index.js file generated by the build.
3. Configure environment variables:
   • GAODE_API_KEY: The API key applied for on the Gaode Open Platform.
   • DINGDING_ACCESS_TOKEN: The access token obtained from the DingTalk robot.

💻 Usage Examples

Basic Usage

1. Select a large model that supports Function Call in Cherry Studio, such as Qwen2.5-7B-Instruct from Guiji Liudong, and complete the corresponding configuration.
2. When the AI receives a weather-related question, it will automatically call the MCP server to query the weather information.
3. After the MCP server obtains the data through the Gaode API, it returns the results to the AI, and the AI organizes the information and replies to the user.

⚠️ Precautions

• Ensure that all environment variables are correct and the third - party API keys are valid.
• Set the custom keyword of the DingTalk robot to "weather" to trigger the message push function normally.
• If you encounter any problems, check the network connection and API call permissions.