Mind Keg

Persistent memory for AI coding agents that stores and retrieves atomic learnings across sessions using semantic search and SQLite.

Provides persistent, cross-session memory for AI coding agents by storing atomic learnings — debugging insights, architectural decisions, and codebase conventions — in a local SQLite database. Supports semantic search via FastEmbed or OpenAI embeddings with keyword fallback, three scoping levels (repository, workspace, global), and API key authentication for multi-repository access control.