MCP server for YouTube video transcription and metadata.
A powerful Model Context Protocol (MCP) server for YouTube video transcription and metadata extraction.
youtube-mcp-server provides MCP tools (get_video_info, transcribe_video) that allow AI agents to retrieve video titles, descriptions, view counts, and full transcriptions without downloading videos, enabling automated content analysis and processing workflows
The server uses in-memory audio processing with yt-dlp extraction, intelligent file-based caching, and parallel segment transcription, eliminating disk I/O overhead while supporting hardware acceleration (CUDA/MPS) for fast inference
The server uses OpenAI Whisper with multilingual support for 99 languages, VAD-based segmentation for precise audio detection, and optional translation capabilities, allowing users to transcribe and translate videos efficiently with configurable model sizes
youtube-mcp-server runs as a local SSE server on port 8000, integrates with MCP clients, and provides JSON-structured outputs for video metadata and transcriptions, making it compatible with existing development tools and CI/CD pipelines
youtube-mcp-server integrates with Model Context Protocol, enabling developers to embed video understanding into AI agents and LLM applications through standardized MCP tools for metadata and transcription retrieval