Project Description

What is Project Description?

Project Description is a Node.js application for Windows OS that serves as a local middleware between LLM-based tools (e.g., Copilot Agents) and local files. It is designed to run alongside an IDE like Visual Studio Code or Visual Studio and uses the @modelcontextprotocol/sdk library. The server includes predefined guide and prompt files for assisting with coding in the Creatio environment.

How to use Project Description?

Install dependencies with npm install, set environment variables GUIDE_FILES_DIR and PROMPT_DIR (either via .env file or terminal), then build with npm run build. For browser testing, run npx @modelcontextprotocol/inspector build/index.js. To integrate with Visual Studio Code, create a .vscode/mcp.json file pointing to the built index.js and follow VS Code’s MCP server documentation.

Key features of Project Description

• Three built‑in tools: ListAvailableFilesTool, ListPromptsTool, ReadLocalFileTool
• Reads guide files (.txt with prefix guide_) and prompt files (.txt with prefix prompt_)
• Uses stdio transport for communication with MCP clients
• Customizable by adding your own guide or prompt files
• Extensible with custom TypeScript tools in src/tools

Use cases of Project Description

• Assist Copilot Agents with local documentation and coding guidelines for Creatio
• Provide structured prompts to control the behavior and output of AI assistants
• Create and manage a local library of reusable coding instructions and examples
• Serve as a lightweight, file‑based knowledge base for LLM tools running in VS Code

FAQ from Project Description

What runtime and OS does Project Description require?

It requires Node.js and is built for Windows OS.

How do I add new guide or prompt files?

Create a .txt file with the prefix guide_ in the folder defined by GUIDE_FILES_DIR, or a .txt file with the prefix prompt_ in the folder defined by PROMPT_DIR.

Where are the data files stored?

They are stored in a predefined ./data directory (guide files) and a ./data/prompts subdirectory (prompt files). The paths are configurable via environment variables.

How can I test the server without an MCP client?

Use the @modelcontextprotocol/inspector tool in a web browser by running npx @modelcontextprotocol/inspector build/index.js.

What transport does the server use and is authentication required?

The server communicates over stdio (as shown in server logs) and no authentication is mentioned in the README.