MemMachine

A self-hosted AI agent memory layer that retains conversations and user preferences across models, with Ollama support and a managed cloud option.

MemMachine gives developers building AI agents a way to remember earlier conversations and user preferences. Its memory persists across sessions, agents and model changes, so user history can carry over when an application switches models.

Conversation recall: Agents can retrieve context from past exchanges and retain short-term context for the current session.
User profiles: Persistent facts and preferences let assistants adapt responses to a user's history without retraining the model for each person.
Model compatibility: It works with Ollama, OpenAI, Anthropic and Bedrock, including models hosted in private environments.
Agent integrations: Connect it to LangChain, LangGraph, CrewAI, LlamaIndex or AWS Strands, as well as applications built with n8n, Dify and FastGPT.
Application access: REST APIs and Python and TypeScript SDKs support custom applications; an MCP server connects to Claude Desktop, Cursor and other compatible clients.

MemMachine is open source under Apache 2.0. You can run its memory server locally, in Docker, on premises or in a private cloud, with memory storage in your own environment. The managed service hosts that server with the provider. Model hosting is a separate choice: self-hosting memory doesn't make calls to cloud model providers local.