McpServerDemo

This MCP server (McpServerDemo) is a lightweight .NET 10 implementation that exposes tools via the Model Context Protocol (MCP) so AI assistants can query and leverage local documents. The server hosts a simple document search tool and demonstrates how to register tools with the MCP framework. You can connect clients like Claude Desktop, Cursor, or VS Code Continue to the server's SSE endpoint to discover available tools and send requests. The server exposes its tools at the SSE endpoint http://localhost:5125/sse once it’s running, and clients can point to that endpoint to start interacting with the tools.

To test locally, you can run the server and then use the MCP Inspector in a browser or a dedicated MCP Explorer client. The README also shows example client configurations for Claude Desktop, Cursor, and VS Code Continue, illustrating how to wire up a local MCP server to those tools via URL and transport settings. Once connected, you can invoke the provided tool (SearchDocs) to perform document-related queries against the local data sources configured in the project.

Prerequisites:

• .NET 10 SDK
• Node.js (for MCP Inspector testing)

1. Install the .NET 10 SDK if you haven’t already:

   • Windows/macOS/Linux: follow the official .NET 10 SDK installation guide for your platform.

2. Obtain the MCP Server Demo source (clone or download the repository):

   • git clone <your-repo-url>
   • cd McpServerDemo/WebApi or navigate to src/McpServerDemo.WebApi as described in the README

3. Build and run the server:

   • cd src/McpServerDemo.WebApi
   • dotnet restore
   • dotnet run
   • The server will start and listen by default on http://localhost:5125 with the SSE endpoint at http://localhost:5125/sse

4. Optional: Test with MCP Inspector (Browser):

   • Ensure Node.js is installed
   • Run: npx @modelcontextprotocol/inspector
   • Open http://localhost:6274 and connect to http://localhost:5125/sse

5. Optional: Test with MCP Explorer (Windows):

   • Download MCP Explorer
   • Connect to http://localhost:5125/sse

6. Stop the server when done (Ctrl+C in the terminal).

Note: The project structure includes the Web API project at src/McpServerDemo.WebApi and a tool named SearchDocs.cs that implements MCP tool behavior. You can extend Tools with new MCP tools by following the pattern shown in the Extending section of the README.

Tips and notes:

• The server hosts an SSE endpoint at /sse. Point MCP clients to this URL to fetch tool metadata and capabilities.
• If you change ports or want to run behind a reverse proxy, you may need to adjust the ASPNETCORE_URLS environment variable or configure Kestrel accordingly.
• To extend functionality, add new tool classes under the Tools folder and annotate them with [McpServerToolType] and [McpServerTool] attributes as demonstrated in the README.
• The included client configuration examples (Claude Desktop, Cursor, VS Code Continue) illustrate how to wire this server into popular AI assistants; adapt the JSON/yaml snippets to your environment if you rename the server or change endpoints.
• Ensure the .NET 10 SDK is installed and that your environment PATH includes dotnet.
• For local testing, you can run the server in one terminal and then run MCP Inspector in another to interactively explore available tools.