Show HN: Memento, Shared Agentic Memory (rcarmo.github.io)

Memento, a new AI tool, has been unveiled to provide a shared agentic memory that outlives individual conversations, catering to users like people, projects, and machines. This system captures vital knowledge—such as chats, reminders, and credentials—while allowing authenticated instances to search, read, and propose adjustments. A two-tiered review process ensures that all changes are vetted by curators before they integrate into the shared repository. Each concept is stored as a Markdown file with stable IDs and rich metadata, while a local Needle model efficiently classifies natural language requests, ensuring that only read operations are executed to maintain data integrity. This innovation is significant for the AI/ML community as it emphasizes collaborative knowledge management across multiple agents without compromising security or accuracy. The layered architecture—incorporating controlled writes, deterministic search, and schema validation—ensures that each interaction remains organized and verifiable. Memento leverages versioned Git storage for diagrams and datasets, enabling a robust environment for maintaining and evolving shared knowledge. By preventing unauthorized modifications and providing a structured proposal process, Memento sets a new standard for cooperation-driven AI systems, paving the way for more sophisticated, memory-driven applications in the realm of artificial intelligence.