MCPTube watches videos so you don't have to. Transcripts, visual frame analysis, and a persistent wiki that gets smarter with every video you add.
Open source ยท MIT License ยท pip install mcptube
Other tools re-discover knowledge from scratch every time. MCPTube is fundamentally different.
Every video you add makes the wiki smarter. Cross-references are built automatically โ connections grow over time, not searched from scratch.
MCPTube doesn't just read โ it watches. Scene-change detection and vision models capture what transcripts miss: diagrams, code, slides, demos.
Works with every leading AI client via the Model Context Protocol. Your video knowledge is available wherever you code and think.
Built on the Karpathy LLM Wiki pattern.
Add a YouTube URL. MCPTube extracts the transcript, detects scene changes, and describes key frames with a vision model.
An LLM extracts structured knowledge and writes wiki pages. Cross-references across videos are created automatically.
Query your knowledge base. The Ask Agent uses FTS5 search and LLM reasoning over compiled knowledge โ not raw chunks.
MCPTube is an MCP server โ it integrates natively with all leading AI clients.
Any MCP-compatible client works out of the box.
We're building the future of video-powered knowledge. Here's what's next.
Managed cloud version โ no setup, no API keys. Just paste a URL and start learning.
Ingest entire playlists and channels. Build domain knowledge bases from curated video collections.
Shared knowledge bases for teams. Learn together, build institutional knowledge from video content.
Visualize your knowledge graph. See connections, gaps, and growth across your video library.
Be the first to try the MCPTube SaaS platform. Join the waitlist for early access and updates.
Join the Waitlist โNo spam. Just updates on launch and early access invites.