xiry.

AI · GitHub repo

memary

Open-source persistent memory layer for autonomous AI agents using knowledge graphs

From the Xiry library · Updated 10 October 2026

Visit github.com ↗

memary is a Python library that equips LLM agents with long-term, human-like memory using knowledge graphs. It tracks user interactions through a Memory Stream and an Entity Knowledge Store, ranking concepts by frequency and recency. By recursively querying and updating graph stores like Neo4j or FalkorDB, it lets agents retain context across sessions.

What it's for

  • Equip autonomous LLM agents with persistent multi-session memory
  • Track user preferences and entity depth over time using knowledge graphs
  • Reduce context token overhead through automated memory summarization and graph retrieval
  • Build multi-user personal assistants using distinct knowledge graph instances

Features

  • Memory Stream to track exposure to concepts with timestamps
  • Entity Knowledge Store ranking concepts by frequency and recency
  • Graph database integration supporting FalkorDB and Neo4j
  • Multi-hop and recursive subgraph retrieval to keep query latency low
  • Local model execution support via Ollama alongside OpenAI endpoints
  • Built-in conversation history summarization to manage context windows
  • Streamlit dashboard for visualizing agent interactions and personas
  • ReAct agent architecture with extensible custom tools

Pricing

Open source

Details

Type
GitHub repo
Category
AI
Topic
Agent Frameworks
Open source
Yes
Platforms
Code library, Linux, macOS, Windows
For
Python developers and AI engineers building stateful LLM agents
Alternative to
MemGPT, Letta, Zep
Tags
agents, memory, knowledge graph, llm, rag, python

Why it was saved

  • @Saboo_Shubham_, 29 December 2024: A drop-in few-lines-of-Python component for adding long-term memory to LLM agent projects.

Related

  • smolagents

    Lightweight open-source Python library for building code-writing and vision AI agents

  • ai-hedge-fund

    Proof-of-concept AI multi-agent trading system with paper trading and backtesting.

  • Eliza

    Open-source multi-agent framework for building autonomous social and on-chain AI agents

  • Self-Improving Agent Loop via skill.md

    A workflow pattern where agents document completed tasks into reusable shortcuts

  • phidata

    Python framework for building multi-modal AI agents with memory and tools

  • AutoGPT

    Open-source platform and framework for building autonomous AI agents

All 5 in Agent Frameworks →

Updated