SimpleMem

SimpleMem is a memory system for developers building AI agents that need to recall past conversations and stored media without passing the full history to a model each time.

• Compact conversation memory: It turns interactions into self-contained facts, resolves references and timestamps, and merges related context to reduce repeated information.
• Targeted recall: Search combines semantic matching, keywords and metadata filters. The system uses the question's intent to select context for the answer.
• Media memory: Python integration provides Omni-SimpleMem's text, image, audio and video memory. A knowledge graph connects information across those formats for questions that draw on multiple kinds of input.
• Retrieval evaluation: EvolveMem evaluates answers on your data, diagnoses failures and proposes retrieval changes, with automatic rollback when evaluation results worsen.

Text memory works through MCP with Claude Desktop, Cursor, LM Studio, Cherry Studio and other compatible clients. The hosted MCP service runs in the cloud; you can instead self-host the server with Docker and keep its memory database on your own host. Full multimodal access uses Python integration.

SimpleMem is open source under the MIT license and requires Python 3.10 or later. Model processing can use Ollama on your machine or an OpenAI-compatible API, including OpenAI, Qwen and Azure OpenAI. Choosing a cloud model sends model requests to that provider.