MemTensor / MemOS

Self-evolving memory OS for LLM & AI Agents: ultra-persistent memory, hybrid-retrieval, and cross-task skill reuse, with 35.24% token savings and DeepSeek Harness support.

Prism-Shadow / penguin-harness

🐧 Unified and Stable RSI Platform

Devin-AXIS / iPolloWork

Enterprise-grade, local-first Agent Workbench for people and agent teams. A unified multi-engine workspace for Codex Harness, DeepSeek Harness, and OpenCode, with unified plugins and Skills, multi-agent projects and tasks, and editable code, documents, presentations, design, and video.

omdsh-dev / dsh-mnemon

Composable, view-based context for DeepSeek Harness. Pluggable sources and strategies, with layered memory and self-evolving skills out of the box.

beita6969 / ScienceClaw

Verifiable program-level self-evolution for AI-for-Science agents: Skills and typed Operators grow from replay-verified executions.

mobius-system / mobius

Self-Evolving Multi-Agent Operation System

TimeLovercc / mojito

Mojito — the local, evolving alternative to cloud personal AI like Meta Muse.