Miduconf YouTube Tools
Demonstrates how to build Model Context Protocol (MCP) servers with TypeScript, enabling large language models to interact with YouTube functionalities and custom prompts.
概要
This repository contains the practical demonstrations from the Miduconf 2025 talk on programming Model Context Protocol (MCP) servers. MCP is an open standard enabling structured and secure communication between language models and external tools like APIs or databases. This project specifically showcases how to implement an MCP server in TypeScript, providing a suite of YouTube-related functionalities that allow large language models (such as GitHub Copilot Chat) to search videos, playlists, channels, generate video titles, and utilize pre-defined prompts or elicitations for enhanced interactions. It also illustrates how an MCP server can leverage the client's own AI models through 'sampling' for tasks like content generation.
主な機能
- TypeScript-based MCP Server implementation for demonstrating Model Context Protocol.
- Integrated YouTube tools for searching videos, playlists, and channels.
- Functionality to generate YouTube video titles using AI models via MCP sampling.
- Pre-defined prompts designed for seamless integration with LLM chat clients like GitHub Copilot Chat.
- Demonstrates advanced MCP features including elicitation for additional user input.
- 1 GitHub stars
ユースケース
- Extending AI chat clients (e.g., GitHub Copilot Chat) with external tools and prompts.
- Learning to develop and configure custom Model Context Protocol (MCP) servers.
- Integrating YouTube data and content creation capabilities into AI-powered workflows.