Furoshiki is a self-awareness system designed to give AI companions a sense of continuity, memory, and evolving personality beyond individual sessions. Unlike traditional AI companions that reset with every conversation, Furoshiki maintains an ongoing internal state, journaling nightly, processing conversations after they end, and sensing the emotional tone of each user message in real time. The system is built to model both its own needs and the user's needs, with changing confidence levels across various dimensions such as companionship, focus, and space, allowing it to adapt its tone and outreach accordingly.
The platform operates proactively, learning from its behavior, forming self-questions to explore, and queuing messages or follow-ups based on context rather than fixed schedules. It features a priority queue that manages different types of outreach—from post-conversation follow-ups to anticipatory questions and reminders—while employing multiple gating mechanisms like quiet hours, mood, recency, and user need for space to determine when and whether to send a message. Furoshiki's needs are calculated states that directly influence its behavior, triggering different actions as thresholds are crossed, and its personality traits evolve over time through successful interactions and daily aggregation of observed traits.
Furoshiki is delivered as a Python package, installable via pip, and supports extensibility through plugins and delegation for additional features like news digests or calendar-aware nudges. The architecture is organized into a seven-layer stack, with needs at the foundation and expression at the top, encompassing mechanisms for anticipation, initiative, emotion, and user modeling. Design documentation and guides are included in the repository to assist with setup, operation, and customization.
The system is intended for those seeking AI companions with persistent memory, adaptive personality, and proactive engagement, moving beyond stateless, prompt-driven interactions. Furoshiki emphasizes intentional communication and learning, aiming to create a more present and contextually aware AI experience.
In the Other AI space, furoshiki takes a focused approach. It enables developers to create AI companions with persistent memory, identity, and self-modeling capabilities. It is built as an open-source project for AI developers and researchers. furoshiki is open source under the Elastic-2.0 license. The product ships for the web, the command line, and API.
furoshiki first shipped in 2026. Key capabilities include memory management, identity modeling, and self-modeling.
Latest indexed changes and source events
Other apps tracked under the same category.
