WRAI.TH transforms multi-agent AI orchestration into an engaging management game where AI agents are robots and projects are planets. Designed for coordinating AI agents, particularly Claude Code, across diverse projects, it provides a robust, local-first platform for agents to register, communicate, remember, and execute tasks. The system features persistent identity for agents, a three-layer knowledge stack for shared memory, and a hierarchical goal cascade for task management, all observable through an intuitive pixel-art galaxy and colony interface. Many of its 58 Model Context Protocol (MCP) tools were even requested and designed by agents themselves, showcasing its unique, agent-driven evolution.
주요 기능
01Flexible multi-agent hierarchy with team-based permissions
02Three-layer knowledge stack: scoped memory, FTS5-indexed vaults, and RAG context
03Persistent agent identity and context restoration across sessions
04Comprehensive messaging with five addressing modes and persistent group conversations
05Goal cascade with prioritized task execution and real-time kanban visualization
063 GitHub stars
