What is aetnamem?

AetnaMem is a Python tool that gives AI agents a long-term memory that stays on your own machine instead of a cloud service. It stores everything in a single SQLite file, and it can install with a simple pip command. The core idea is that every fact an agent remembers, and every action it takes, leaves a trail you can inspect and verify later, rather than trusting the system blindly. At its simplest, AetnaMem lets an assistant remember, recall, correct, or forget facts about a user across separate conversations. It works with different agent frameworks and model providers, including Claude, Grok, OpenAI, DeepSeek, and Ollama, because the memory itself is kept separate from whichever AI model is answering. There is also a plugin for a tool called OpenClaw that adds this memory automatically, and support for connecting through a shared protocol called MCP so other agent tools can discover and use it. Beyond plain memory, AetnaMem adds an optional layer for guarded actions: before an agent is allowed to change something in the real world, the exact planned operation can be staged, tied to a human approval, rechecked against the current state, and only then carried out, with a receipt produced afterward. Content pulled from webpages or other tools is kept quarantined and not treated as trusted memory until someone explicitly approves it. There is also a structured decision workflow, useful for teams like a hospital or policy group who need panels of people to review evidence and vote on a recommendation, with every step logged and separately checked. AetnaMem is licensed under AGPL 3.0, and the project includes benchmark results, a scientific report, and a desktop-style dashboard for chatting locally with your memory and approvals visible. It is aimed less at casual chatbot use and more at people building agent systems who need to prove, after the fact, what an agent knew and why it acted. The full README is longer than what was shown.