Agent Memory Infrastructure
13.3k 2026-04-13
MemoriLabs/Memori
Memori provides SQL-native, LLM-agnostic memory infrastructure for AI agents, transforming execution and conversation into structured, persistent state.
Core Features
SQL-native persistent memory for AI agents
LLM-agnostic integration with various large language models
Transforms agent execution and conversation into structured data
Seamless integration with existing software and infrastructure
Proven high accuracy and token efficiency (LoCoMo benchmark)
Quick Start
pip install memoriDetailed Introduction
Memori is a crucial infrastructure layer designed to provide robust, persistent memory for AI agents. Unlike traditional methods that rely solely on conversational context, Memori captures agent actions and dialogues, structuring them into a SQL-native format. This approach ensures that agents retain long-term memory, improving performance and reducing token costs by recalling relevant information efficiently. It's LLM and datastore agnostic, allowing flexible integration into diverse production systems, making it ideal for building production-ready AI applications.