mem0 and Memary
These are direct competitors both targeting the same problem of persistent memory infrastructure for autonomous agents, with Mem0 being the more mature and widely-adopted option while Memary offers a lighter-weight alternative.
About mem0
mem0ai/mem0
Universal memory layer for AI Agents
Implements multi-level memory (user, session, agent state) with adaptive retrieval that achieves 26% higher accuracy and 90% lower token usage than baseline approaches. Supports multiple LLMs and vector stores, with SDKs for Python and JavaScript, plus integrations for LangGraph and CrewAI. Offers both self-hosted open-source deployment and a managed platform with CLI tooling for memory management operations.
About Memary
kingjulio8238/Memary
The Open Source Memory Layer For Autonomous Agents
Implements a multi-layered memory architecture combining episodic memory streams, entity knowledge graphs (via FalkorDB or Neo4j), and dynamic user/system personas that automatically evolve through agent interactions. Supports both local models via Ollama (Llama 3, LLaVA) and OpenAI APIs with pluggable tools, enabling developers to integrate memory into existing LlamaIndex ReAct agents or use the built-in ChatAgent implementation with minimal code changes.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work