MemoryOS

Active
GitHub Python Apache-2.0

Description

[EMNLP 2025 Oral] MemoryOS is designed to provide a memory operating system for personalized AI agents.

Key Features

  • Hierarchical memory architecture (short-term, mid-term, long-term) inspired by OS memory management
  • MCP Server integration for injecting long-term memory into any AI application
  • SOTA performance on LoCoMo benchmark with 49.11% F1 and 46.18% BLEU-1 improvement
  • Plug-and-play modular design supporting pluggable storage engines, update strategies, and retrieval algorithms
  • Universal LLM support including OpenAI, DeepSeek, Qwen, and Anthropic models
  • Docker deployment and ChromaDB vector database support for scalable memory storage

Use Cases

💡 Enable personalized AI agents with long-term conversational memory and user profiling
💡 Build context-aware chatbots that remember user preferences across sessions
💡 Create RAG-powered applications with dynamic memory updating and retrieval
💡 Develop AI coding assistants that recall previous project context and decisions

Tags

agent-memory rag personalization python

Categories

🧩 Agent Memory 📚 RAG Tools

Quick Start

1. Install via pip: `pip install memoryos` or via MCP: configure in your agent client.
2. Set environment variables for your LLM provider (e.g., `OPENAI_API_KEY`).
3. For MCP integration: add MemoryOS MCP server to your Claude Desktop or Cursor config.
4. Use the Playground platform at baijia.online/memoryos/ for quick testing.
5. For Docker deployment: use the provided docker-compose setup with ChromaDB.
Visit GitHub

Related Projects