LEANN is a lightweight personal vector database that achieves 97% storage savings compared to traditional solutions while maintaining search accuracy. Published at MLsys 2026, it uses graph-based selective recomputation with high-degree preserving pruning to compute embeddings on-demand instead of storing them all. This allows indexing 60 million text chunks in just 6GB instead of 201GB. LEANN can semantic search across local files, emails, browser history, chat logs (WeChat, iMessage), AI conversations (ChatGPT, Claude), live data (Slack, Twitter), and codebases with intelligent language-aware chunking. Native MCP integration makes it a drop-in semantic search service for Claude Code. All data stays local with zero cloud costs, ensuring complete privacy.