Tolaria is a desktop application developed for macOS and Linux, designed for the management of Markdown-based knowledge bases. It serves as a tool for various use cases, including personal knowledge management (second brains), organization of company documentation for AI contexts, and the structured storage of AI agent memories and operational procedures. The application is specifically engineered to handle large personal vaults, with the developer using it for workspaces exceeding 10,000 notes.

The core methodology of Tolaria is built upon a set of fundamental principles that prioritize user data ownership, flexibility, and integration with standard developer tools:

1. Files-First Paradigm: All notes are treated as plain Markdown files. This design ensures inherent data portability, compatibility with any standard text editor, and eliminates the need for proprietary export processes. The user retains full control and ownership of their data, independent of the application.
2. Git-First Architecture: Each knowledge vault within Tolaria is fundamentally a Git repository. This architectural choice provides comprehensive version history for all content, enables seamless synchronization with any standard Git remote, and removes any dependency on Tolaria-specific servers for data management or version control. This leverages Git's distributed version control system capabilities.
3. Offline-First and Zero Lock-in: The application operates completely offline, without requiring accounts, subscriptions, or cloud dependencies. This ensures uninterrupted access to the vault regardless of internet connectivity and guarantees that user data remains accessible and usable even if the user discontinues using Tolaria.
4. Open Source Model: Tolaria is distributed as free and open-source software, encouraging transparency, community contributions, and broader accessibility.
5. Standards-Based Data Formats: Notes are Markdown files, optionally augmented with YAML frontmatter for metadata. This adherence to widely adopted open standards prevents vendor lock-in, ensuring that data is interoperable with a vast ecosystem of existing tools and workflows.
6. Types as Lenses, Not Schemas: Tolaria implements "types" not as rigid data schemas with enforcement mechanisms like required fields or validation rules, but rather as flexible navigation aids. These types serve as helpful categorization labels to facilitate the discovery and organization of notes.
7. AI-First but AI-Optional Integration: While designed to interoperate effectively with AI agents, Tolaria does not impose AI usage. It facilitates AI interaction by supporting setup paths for common AI models like Claude Code, Codex CLI, and Gemini CLI, and provides an AGENTS.md file intended for AI agent guidance. For Gemini CLI specifically, Tolaria can register its internal MCP (Master Control Program) server in ~/.gemini/settings.json, scoped to the active vault's path, and offers a GEMINI.md compatibility shim to redirect Gemini to the shared AGENTS.md instructions non-destructively.
8. Keyboard-First Design: The user interface, particularly the editor and command palette, is optimized for keyboard navigation and command execution, catering to power users who prefer keyboard-driven workflows.
9. Real-World Use Driven Development: The feature set of Tolaria is a direct result of solving practical problems encountered during the management of the developer's personal, extensive knowledge base, ensuring functionality is grounded in utility.

Technically, Tolaria is built using the Tauri framework for cross-platform desktop application development, leveraging React for its user interface and TypeScript for type-safe JavaScript development. Local development requires Node.js 20+, pnpm 8+, and Rust stable, with specific system dependencies for Linux environments (e.g., WebKit2GTK 4.1, GTK 3, build-essential packages). The bundled MCP server relies on the system's Node.js binary at runtime for external AI tooling on Linux. Architectural details, core abstractions, and codebase navigation guides are provided through dedicated documentation files (ARCHITECTURE.md, ABSTRACTIONS.md, GETTING-STARTED.md, and Architecture Decision Records). The project is licensed under the AGPL-3.0-or-later.