A new open-source knowledge management tool called Atomic transforms markdown notes into a semantically-connected knowledge graph with AI-powered features. The platform stores information as 'atoms' — markdown notes that are automatically chunked, embedded, tagged, and linked by semantic similarity. Users can synthesize atoms into wiki articles, explore connections on a spatial canvas, and query their knowledge through an agentic chat interface.

Built with Rust and React, Atomic runs as a desktop app using Tauri, a headless server with Docker, or a native iOS client. The system supports multiple AI providers including OpenRouter for cloud-based embeddings and Ollama for local processing. Key features include semantic search using sqlite-vec, force-directed spatial visualization, LLM-generated wiki synthesis with inline citations, and a browser extension for capturing web content directly into the knowledge base.

Developers can self-host Atomic using Docker Compose or deploy to Fly.io with persistent storage. The project architecture separates core logic into a standalone Rust crate, with clients acting as thin wrappers. An MCP server allows integration with Claude and other AI tools, while hierarchical tagging and RSS feed support provide additional organizational capabilities. Atomic is licensed under MIT and available on GitHub for developers seeking a self-hosted alternative to commercial knowledge management platforms.