Memori Labs 🇺🇸

About the channel

Memori is agent-native memory infra: an LLM-agnostic layer that transforms agent execution and conversation history into structured, persistent state for production systems. Unlike conversational memory wrappers or vector retrieval, Memori captures tool calls, decisions, traces, and user context as durable, queryable state that agents rely on across sessions, models, and workflows. Memori is benchmarked on LoCoMo, an independent benchmark from Snap Research measuring how well systems remember and retrieve information. Memori delivers industry-leading results: 87% accuracy using only 721 tokens, 2.8% of full-context cost, enabling teams to reduce inference spend by up to 97.2%. Built for the enterprise, Memori works with the data infra you already run through partnerships with database providers, no rip-and-replace, and deploys across managed cloud, single-tenant cloud, VPC, and on-premises. Memori’s open-source repository has surpassed 15.6K GitHub stars and 550K downloads.