Tolaria is a new open-source macOS app for managing markdown knowledge bases. It ships with built-in support for AI agents like Claude Code and Codex CLI. Created by Luca Ronin, the app includes an AGENTS file in each vault that gives these agents persistent context about how to interact with your notes. Ronin uses it daily to manage a personal vault of 10,000+ notes. The AGENTS file is a configuration layer that makes your markdown legible to machines. It defines agent memory and procedures. It covers which folders matter and what operations are common. It lays out how the project is structured. When Claude Code or Codex CLI opens your vault, they parse this file to understand what they're working with. Agents get stateful interactions that persist across multiple sessions. For anyone building agent tooling, this convention-over-configuration pattern is a clean way to bridge human knowledge bases and AI agents.
The rest of Tolaria's design follows the same philosophy of user control. Notes are plain markdown files with YAML frontmatter. Every vault is a git repository, so you get full version history without depending on any cloud service. No accounts or subscriptions. No vendor lock-in. Built with Tauri, React, and TypeScript, the app is licensed under AGPL-3.0-or-later.
The approach is simple. Don't build complex integrations. Give agents a readme they can parse. In practice, that means dropping a YAML file in your repo root that tells agents where to find specific content and how to handle it. Agents read it once and carry that understanding forward. Ronin runs his 10,000+ note vault this way daily.