Kernal Mcp

This is a local-first knowledge graph that treats conversation as structured data. It exposes 13 MCP tools for Claude to extract people, organizations, activities, and action items from natural language and store them in a local SQLite database. The pattern is clever: you tell Claude about a meeting, it calls kernal_remember with the raw text, then Claude decides what entities to extract and calls structured write tools like kernal_add_person and kernal_add_org. Entity resolution with fuzzy matching prevents duplicates. Query tools like kernal_context give you full briefings on contacts and companies, kernal_actions surfaces follow-ups, and kernal_recall searches everything. Reach for this if you want to build a personal CRM from conversational updates without sending data to a third party service.