AI assistants waste tokens loading context they don't need. Hive queries your Obsidian vault on demand via MCP — only what's relevant, when it's relevant. 67-82% token reduction.