YggdrasilBAI-LAB/MemoryOS
↗ GitHub[EMNLP 2025 Oral] MemoryOS is designed to provide a memory operating system for personalized AI agents.
1,302
Stars
128
Forks
9
Watchers
14
Open Issues
Safety Rating A
The repository is a legitimate academic research project backed by a published EMNLP 2025 paper. No hardcoded secrets, malicious code patterns, suspicious dependencies, or prompt injection attempts were found. API keys are handled via user-supplied configuration. The codebase follows standard Python packaging and dependency practices.
ℹAI-assisted review, not a professional security audit.
AI Analysis
MemoryOS is a hierarchical memory management system for personalized AI agents, inspired by operating system memory management principles. It provides short-term, mid-term, and long-term memory layers with automated storage, updating, retrieval, and generation modules. Accepted as an EMNLP 2025 Oral paper, it achieves state-of-the-art results on the LoCoMo benchmark with 49.11% and 46.18% improvements in F1 and BLEU-1 scores respectively. It supports MCP (Model Context Protocol) server integration, ChromaDB vector storage, Docker deployment, and multiple LLM providers including OpenAI, Anthropic, DeepSeek, and Qwen.
Use Cases
- Adding persistent long-term memory to AI agents and chatbots
- Enabling personalized AI interactions by maintaining user profiles and knowledge across sessions
- Integrating memory capabilities into agent clients like Claude Desktop, Cline, and Cursor via MCP
- Building RAG pipelines augmented with hierarchical personal memory
- Benchmarking and evaluating AI memory systems on the LoCoMo dataset
Tags
Security Findings (1)
No hardcoded secrets detected. The README and code samples use placeholder values such as 'YOUR_OPENAI_API_KEY' and instruct users to supply their own API keys via configuration files or environment variables.