MemMachine

MemMachine is an open source memory layer for AI agents. It lets applications store, recall and learn from data and preferences gathered in past sessions, so assistants can respond with context instead of starting every conversation from zero. The memory persists across sessions, agents and language models, and builds an evolving user profile. The system provides short term, long term and personalization memory, and is model agnostic, so one memory store can serve several models at once, including specialized models hosted in a private cloud or an on premises data center. It is aimed at developers and organizations building AI applications who want to keep control of their data rather than relying on a single model provider's built in memory. The vendor describes the underlying approach as combining vector storage with graph based entity and temporal relationships. The project was incubated by MemVerge and is now community driven. Pricing is published in three tiers. The Open Source tier costs nothing and is self hosted on your own infrastructure, with all memory types, core integrations and community support. The Pro tier costs $79 per month, billed per project, and adds managed cloud hosting, higher rate limits, premium integrations and email and chat support. Yearly billing is offered with a stated 15 percent saving. The Enterprise tier is custom priced and adds on premises or VPC deployment, custom service agreements, SSO and advanced security, and a dedicated engineer. A free sandbox is offered for trying the product.