This YouTube Transcript MCP server acts as a specialized bridge between video content and AI assistants like Claude or Cursor, allowing them to "read" videos instantly. In simple terms, it automatically retrieves the written text from any YouTube video so the AI can analyze it without a human needing to manually watch the video or copy-paste subtitles. This makes it incredibly easy for anyone to ask an AI to summarize a long tutorial, extract key points from a lecture, or find specific information hidden within hours of footage. Beyond simple text retrieval, the server provides five distinct tools that offer a high degree of control over how video data is processed. It can fetch transcripts for single videos or handle batch processing for multiple links at once. The system also includes capabilities for translating transcripts into different languages and reformatting the output into specific files like SRT or VTT. This flexibility ensures that AI models receive exactly the data they need in the most useful structure possible for generating summaries or structured reports. For developers building sophisticated AI workflows, this Go-based implementation is engineered for high performance and production reliability. It solves common data-fetching challenges by incorporating built-in rate limiting and proxy rotation to navigate YouTube's access restrictions. To ensure lightning-fast response times, the server supports both in-memory and Redis caching, which prevents redundant requests for popular videos. Fully compliant with the latest Model Context Protocol standards and Docker-ready, it provides a robust, scalable infrastructure for any application requiring deep video context.
Category: Design, Media & Creative
Tags: subtitles, summarization, transcription, video, youtube
Visit YouTube Transcript MCP Server
bash # Clone the repository git clone https://github.com/kyong0612/youtube-mcp.git cd youtube-mcp # Run the installer ./scripts/install-mcp.sh Via Go Install bash # Install the stdio version for MCP clients go install github.com/kyong0612/youtube-mcp/cmd/mcp@latest # Rename the binary for clarity (standard Go path is usually $GOPATH/bin) mv $GOPATH/bin/mcp $GOPATH/bin/youtube-mcp-stdio Manual Build bash # Build the stdio server directly from the source go build -o youtube-mcp-stdio ./cmd/mcp/claude_desktop_config.json or Cursor's MCP settings): json { "mcpServers": { "youtube-transcript": { "command": "/path/to/youtube-mcp/youtube-mcp-stdio", "args": [], "env": { "LOG_LEVEL": "info", "CACHE_ENABLED": "true", "YOUTUBE_DEFAULT_LANGUAGES": "en,ja" } } } } Note: Ensure the command path points to the actual location of your youtube-mcp-stdio binary. Key Environment Variables: - YOUTUBE_DEFAULT_LANGUAGES: Default languages for transcripts (e.g., "en,ja"). - CACHE_TYPE: Type of cache to use (memory or redis). - LOG_LEVEL: Logging level (debug, info, warn, error). - CACHE_ENABLED: Set to true to enable caching.get_transcript: Fetch the transcript for a single video. * Arguments: video_identifier (URL or ID), languages (list), preserve_formatting (boolean). * get_multiple_transcripts: Batch process transcripts for multiple videos. * Arguments: video_identifiers (list), languages (list), continue_on_error (boolean). * translate_transcript: Translate transcripts into different languages. * format_transcript: Format transcripts into specific styles (plain text, SRT, VTT, etc.). * list_available_languages: List all available subtitle languages for a specific video.format_transcript tool to clean up the text and the get_transcript tool to provide the raw data needed to generate high-quality written content without the creator needing to leave their workspace. Example: A YouTuber provides a video link to Claude and says: "Fetch the transcript for this 30-minute video, summarize the three most important takeaways, and draft a 500-word blog post based on the speaker's exact tone."translate_transcript and list_available_languages tools, a user can fetch a foreign-language transcript and have it translated into their native language. Because the transcript is pulled into the AI’s context window, the user can ask complex follow-up questions about specific technical concepts mentioned in the video. Example: A developer uses the MCP in Cursor to pull the transcript of a Japanese developer keynote, translates it to English, and then asks: "Based on this transcript, what are the three breaking changes mentioned for the new API version?"get_multiple_transcripts tool allows for batch processing. An analyst can provide a list of URLs to the AI, which then fetches all transcripts simultaneously. The AI can then perform a horizontal analysis across all videos to find common themes, pros, and cons. Example: "Fetch the transcripts for these five different 'iPhone 16 Pro Review' videos. Create a table comparing what each reviewer said about the new 'Camera Control' button, specifically noting if they found it intuitive or gimmicky."VTT Generation) Problem: Educators or developers building…
Part of MCP Servers
YouTube Transcript MCP Server allows AI assistants to fetch subtitle transcripts from YouTube videos using five distinct tools. It can fetch single transcripts, execute batch queries across multiple video links, list supported subtitle languages, convert output into plain text, SRT, or VTT formats, and retrieve YouTube-provided translated captions.
You can install it with Go 1.24 or higher by running go install github.com/kyong0612/youtube-mcp/cmd/mcp@latest and renaming the generated binary to youtube-mcp-stdio. Alternatively, you can clone the source repository from GitHub and execute the automated setup script ./scripts/install-mcp.sh.
The server works with any Model Context Protocol client supporting stdio transports, including Claude Desktop, Claude Code, and Cursor. It also offers an optional HTTP server mode that allows developers to run tool calls over standard JSON-RPC HTTP requests.
Yes, YouTube Transcript MCP Server is open source and licensed under the MIT License. The code repository, documentation, and Docker Compose configurations are publicly available on GitHub at https://github.com/kyong0612/youtube-mcp.
No, the translate_transcript tool fetches captions in the specified language, including YouTube's auto-translated captions when they are already available on the platform. It does not perform custom machine translation of arbitrary text on its own.