xuiltul/animaworks

Implements autonomous agent orchestration using isolated OS processes (one per agent) with inter-process communication and automatic restart, neuroscience-inspired memory architecture combining RAG with three-stage forgetting and multi-channel priming, and supports six LLM engines (Claude, GPT, Gemini, Codex, LiteLLM, and local models). Agents operate continuously via configurable heartbeat cycles and cron scheduling rather than request-response workflows, coordinate through message passing on shared channels, and are visualized in real-time dashboards including a 3D office environment with spatial presence.

159 stars.