Memanto; open-source memory agent that remembers, recalls and answers (github.com)

Memanto has launched as an innovative open-source memory agent designed for AI systems, enabling persistent recall and context management for agents including Claude Code, Codex, and Cursor. Unlike traditional memory solutions that act as passive infrastructures, Memanto is an active memory agent that allows AI to genuinely remember, recall, and answer queries based on its learned context, enhancing long-term goal execution and reducing confusion. It operates locally on user machines via Docker or as a free cloud service, requiring no API keys or complex backend systems, which makes it easily accessible for developers. The significance of Memanto lies in its advanced capabilities that address existing memory gaps identified by AI models themselves, such as eliminating static memory injection and providing versioning and temporal signals for awareness of context changes. Its retrieval system leverages the Moorcheh semantic engine for immediate, exact memory access without the delays associated with indexing typical of vector databases. By categorizing memories into 13 types—such as facts, goals, and context—Memanto facilitates more efficient searches and interactions by agents, promising reduced token usage and enhanced information management capabilities in AI workflows.