P2P desktop app for building, browsing, & sharing knowledge
a local-first AI workspace a mail client a personal organizer for time and events a decentralized knowledge and message sharing platform a remote interface for reasoning over your work through Telegram with local models The project exists because I wanted a better way to share research with other people. It is open source and still in progress.
Subgrapher includes an AI agent that reasons over references, notes, mail, and files within the local-first workspace. Users can attach folders, mail threads, and web content to references, then leverage AI to analyze and synthesize information without sending data to external platforms.
Subgrapher supports both private peer sharing and public publishing of semantic references. Users can selectively share knowledge containers with specific trusted contacts or make them publicly available, enabling decentralized knowledge exchange without relying on centralized platforms.
Subgrapher functions as a local-first AI workspace supporting both local and API-based models, allowing users to work without internet connectivity and maintain data sovereignty. The desktop app architecture ensures users retain control over their information and can choose their preferred AI infrastructure.
Subgrapher lets users create semantic references as reusable knowledge containers that can be forked, shared privately with peers, or published publicly. This enables researchers and knowledge workers to organize complex information across files, messages, and web content while maintaining ownership and portability of their work.