The project architecture centers on a core Rust library (atomic-core) with thin clients including a desktop UI, server, and MCP, plus a multi-client frontend (React) that runs across platforms.

AI provisioning can connect to either cloud or local models, requiring an API key or local setup via OpenRouter (cloud) or Ollama (local).

Core features encompass semantic search with sqlite-vec, a force-directed canvas visualization, wiki synthesis with inline citations, a chat interface for knowledge-base conversations, auto-tagging, multiple AI providers (OpenRouter or Ollama), RSS feeds, a browser extension, an MCP server for Claude and others, multi-database support, and an iOS app.

Getting started includes desktop app delivery via Tauri and a headless server option, with detailed instructions for desktop download, Docker Compose self-hosting, Fly.io deployment, and standalone server setup.

A browser extension enables web clip capture into Atomic, and the MCP server exposes endpoints for external tools to perform search and atom creation.

Prerequisites include Node.js 22+, a Rust toolchain, and platform-specific dependencies for the Tauri-based desktop application.

Development workflows cover build, development, and tests across frontend and backend, with a tech stack spanning Rust, SQLite with sqlite-vec, Actix-web, React, Vite, Tailwind, and iOS SwiftUI; licensed under MIT.

Atoms are markdown notes automatically chunked, tagged, embedded, and linked by semantic similarity; they can be synthesized into wiki articles, explored on a spatial canvas, and queried via an agentic chat interface.

Atomic is a personal knowledge base that converts markdown notes into a semantically-connected, AI-augmented knowledge graph.

Project structure includes crates for core logic, server, MCP, and multiple frontends (desktop, web, iOS), plus a browser extension and related extension scripts.