Open-source AI memory engine for agents: multimodal memory, temporal knowledge graphs, graph RAG, and evidence-backed long-term context.

A human-in-the-loop Web Agent that never makes the same mistake twice. MemBrowser combines VLM, DOM stripping, and multi-modal memory to turn brittle AI interactions into robust, zero-cost RPA workflows.一个“教一次就会”的混合架构 Web Agent。结合人类示教（Human-in-the-loop）与多模态记忆网络，彻底解决大模型浏览器自动化的幻觉与高昂成本。