Project Description

This project is a Node.js application for Windows OS called MCP Server. It serves as a local middleware for seamless communication between LLM-based tools (e.g., Copilot Agents). The server is designed to be run locally, ideally alongside an IDE like Visual Studio Code or Visual Studio.

MCP Server uses the third-party library @modelcontextprotocol/sdk, which allows you to run and test it in a web browser using the Inspector tool.

The project includes a predefined ./data directory that contains documentation and guide files used to assist with coding. Additionally, a ./data/prompts subfolder contains prompt files used by Copilot Agents to guide their behavior and output.

Configuration

Set the following environment variables:

GUIDE_FILES_DIR: Path to the directory containing your local guide files (e.g., .\\data).
PROMPT_DIR: Path to your prompt directory (e.g., .\\dat\\prompts). Prompt files must be .txt files and start with the prefix prompt_.

You can set these variables in a .env file or pass them directly in the terminal when running the app. Keep in mind having dobule "\" in your path names.

Example

GUIDE_FILES_DIR=C:\\pathToYourProject\\mcpLocalServer\\data\ PROMPT_DIR=C:\\pathToYourProject\\mcpLocalServer\\data\\prompts

💡 The paths should directly point to the folders where your guide and prompt files are located. It’s recommended to create a prompts subfolder under data.

Related MCP server: interactive-mcp

Features

There are 3 build-in tools which you can use:

ListAvailableFilesTool – Lists all .txt files located in the data directory.
ListPromptsTool – Lists all prompt files located in the data/prompts directory.
ReadLocalFileTool – Reads and returns the content of a file from either data or data/prompts, depending on the file prefix.

Usage

Install dependencies:
npm install
Configure environment variables using a .env file or directly in the terminal.
Build the project:
npm run build
Run the server for browser testing:
npx @modelcontextprotocol/inspector build/index.js
Use the MCP-compatible client (like Copilot for Visual Studio Code) to interact with the server.

Usage with Visual Studio Code

To integrate with Copilot:

Create a .vscode directory in your project.
Add a mcp.json file with the following content:

{ "inputs": [], "servers": { "local": { "command": "node", "args": ["C:\\pathToMcpLocalServerProjectInYourPC\\build\\index.js"] } } }

Follow the instructions from the official documentation to enable MCP support in VS Code.
Start your server if is disabled with a command: ctrl+shift+p MCP:List Servers -> next chose you server Name and click action [""Start","Stop","Restart"]
You should see MCP server Logs in OUTPUT terminal: 2025-05-30 08:35:10.855 [warning] [server stderr] ✅ Local MCP server running on stdio 2025-05-30 08:35:10.864 [info] Discovered 3 tools

⚠️ Make sure you're using the latest or pre-release version of Visual Studio Code that supports Agents.

Example prompt using guide files and custom prompts: You can find more examples in "../docs/VisualCode_Copilot_Examples"

Description: 💾 Create C# Class Prompt text: Use the local MCP server. Step 1: Load the prompt file named "prompt_csharp_class_authoring.txt" using the "read_local_file" tool with the argument "fileName". Step 2: Once the file is loaded, execute all instructions defined in that file. If those instructions require using additional tools, follow them accordingly. Step 3: After completing all steps from the prompt file, generate a C# (.cs) file with a class named House. The class should contain Windows and Doors properties, and include methods to build them. Ensure that the implementation follows all guidelines and conventions from the loaded prompt file.

Customization

You can extend this project by adding your own guide or prompt files:

Add a guide file: Create a .txt file with the prefix guide_ and place it in the folder defined by GUIDE_FILES_DIR.
Add a prompt file: Create a .txt file with the prefix prompt_ and place it in the folder defined by PROMPT_DIR.

You can also develop custom tools: