Zep

Long-term memory and data persistence for AI agents and LLM applications.

Zep provides the memory layer for production AI (built for speed with a C++ core). It handles session history, automated summarization, and vector embeddings to ensure LLMs maintain context over weeks, not just minutes. The platform reduces token waste by pruning irrelevant data while surfacing key facts through a searchable memory store. It integrates via Python or TypeScript SDKs and supports local deployment or cloud hosting for sub-10ms retrieval.

https://www.getzep.com