peterfei/ai-agent-team

Implements a vector-based memory system with semantic search using a built-in embedding model (Xenova/all-MiniLM-L6-v2), enabling AI agents to retain and recall context across sessions with 30-50% improved accuracy compared to time-based retrieval. Operates as a Model Context Protocol (MCP) server integrated with Claude Code, managing persistent SQLite-backed conversation threads with automatic Git version control and role-specific agent initialization commands. Supports both global and project-local configurations with zero-download installation, bundling all dependencies for offline-first operation in restricted network environments.

290 stars. Available on npm.