SimpleMem
Open-source AI agent memory with Docker self-hosting and Ollama support for text recall. Python integration adds image, audio and video memory. MIT licensed.
SimpleMem is a Python memory system for developers building AI agents that need to recall earlier conversations and stored media. You can self-host it with Docker or use its hosted text-memory service. The code is open source under the MIT license.
Its main focus is reducing how much history an agent has to process. It turns conversations into compact facts, resolves references and dates, and merges related information as it stores it. Retrieval combines semantic search, keyword matching and metadata filters, with the query's intent guiding which memories reach the model.
Text memory connects to MCP clients such as Claude Desktop, Cursor, LM Studio and Cherry Studio. Python integration also provides access to Omni-SimpleMem, which handles images, audio and video alongside text. Its knowledge graph connects information across those media so an agent can answer questions that draw on more than one type of input.
Storage can stay local. The self-hosted server keeps its memory database on the host, while model processing uses Ollama or an OpenAI-compatible API. Choosing a cloud provider such as OpenAI, Azure OpenAI or Atlas Cloud sends model requests to that service; the hosted MCP offering also runs outside your machine.
For developers evaluating recall on their own data, EvolveMem can diagnose failed answers and propose changes to retrieval settings. It checks those changes against evaluation questions and rolls them back when results worsen.
