Markdown notes that treat Git and agents as first-class citizens

Tolaria is a desktop app for people who already trust plain Markdown, but want a working surface that understands a large vault. The README frames it as a manager for Markdown knowledge bases on macOS, Windows, and Linux. That is accurate, but undersells the more interesting angle: Tolaria is trying to make a folder of notes usable as a structured knowledge graph for both a human and a local AI agent.

The basic model is conservative. A vault is a folder on disk. Notes are .md files with YAML frontmatter. The app reads and writes the filesystem rather than hiding your data in a hosted backend. Git is recommended for history, recovery, sync, and remote collaboration, but the docs say a plain folder still works. That makes Tolaria closer to Obsidian, Logseq, or a custom notes repo than to a SaaS workspace.

What makes it worth watching in June 2026 is the agent layer. Tolaria can generate setup paths for Claude Code, Codex CLI, and Gemini CLI, and its own docs describe the vault as a graph of people, projects, events, responsibilities, and ideas. The goal is a local corpus that an agent can navigate without needing a private export format, with prettier note-taking as a side effect.

What it actually gives you

Tolaria’s strongest decisions are about ownership and recoverability:

  • Notes, attachments, type definitions, saved views, and Git metadata live in the vault folder.
  • App-only preferences such as window size, zoom, recent vaults, and cache data stay on the local machine.
  • Markdown plus YAML frontmatter is the storage format, with conventions for fields such as type, status, belongs_to, related_to, and has.
  • Multiple vaults can be mounted into one graph, while commits and Git sync remain tied to each vault’s own repository.
  • The app is built with Tauri, React, TypeScript, and Rust, so it ships as a native desktop app rather than a browser-only notes UI.

This is the part README-style summaries often miss: Tolaria is opinionated about graph structure, but it does not require a database. Types and relationships are conventions layered onto files. If the app disappears, the notes are still there. If an agent needs context, it can read the same files.

Install

The documented macOS path is Homebrew:

brew install --cask tolaria

The project also publishes builds for macOS, Windows, and Linux from its download page and GitHub releases. Its install docs mark macOS as the primary day-to-day development target, while Windows and Linux are supported but early. Windows installers may still need approval on company-managed devices, and Linux desktop behavior depends on distribution WebKitGTK and input-method integration.

For local development, the README gives this path:

pnpm install
pnpm dev
pnpm tauri dev

Linux development needs WebKit2GTK 4.1 and GTK 3 dependencies. The README lists package commands for Arch, Debian or Ubuntu 22.04+, and Fedora 38+.

The workflow it is built around

The first launch path makes the storage model explicit. You can clone a Getting Started vault, open an existing folder of Markdown files, or create a new empty vault. The sample vault is cloned locally and then disconnected from its remote, so tutorial edits do not accidentally push upstream.

Once inside, Tolaria treats a vault as more than a file tree. It scans Markdown files, reads frontmatter, builds note lists, folders, saved views, backlinks, relationships, and search surfaces. The command palette is the main control plane. Git support adds diffs, commits, sync, history, and recovery if the vault is a repository.

Framing it as “notes app versus IDE” misses the mark. Tolaria sits between them. It gives a writer or operator a keyboard-heavy desktop app, while keeping the same folder legible to Claude Code or Codex. It targets a personal operating manual, research vault, team runbook, or agent memory store.

Where it is still early

The project is young: the repository was created in February 2026, and the latest visible releases at the time of writing are alpha builds from June 13, 2026. It has roughly 16,000 stars and 52 open issues as of 2026-06, so attention has arrived before the cross-platform polish is fully settled.

The issue tracker shows the kinds of rough edges you would expect in a Markdown desktop app that spans platforms and local AI tooling. Recent open issues mention Windows path-style wikilinks, Linux images in subdirectories, macOS refresh behavior, Chinese punctuation input, AI chat wrapping, and Codex CLI chat failing on Windows when PATH resolves through a Volta shim. None of those invalidates the architecture, but they matter if you want to use Tolaria as a daily workspace today.

The license also deserves a direct read. Tolaria is AGPL-3.0-or-later. That is normal for a project trying to keep improvements open, but it is not the same operational choice as MIT. Teams embedding or modifying it should route that through their normal open-source review.

Tolaria versus the obvious alternatives

Obsidian is still the safer pick if you want a mature plugin ecosystem, polished mobile support, and a huge community around personal knowledge management. Logseq is closer if outlines and block-level thinking are your default. A bare Markdown repo is cleaner if you only need files and Git.

Tolaria’s argument is narrower and sharper: a native Markdown vault app where Git, frontmatter conventions, mounted workspaces, and AI-agent setup are part of the core model. It does not chase the most extensible notes universe. Its goal is to make a local knowledge base operational.

That trade makes sense if your notes are also working context. If you want your agent to understand projects, procedures, people, links, and saved views from the same files you edit every day, Tolaria is more interesting than another Markdown editor. If you mainly want polished consumer note-taking, wait for the alpha edges to come down.

Star growth

The sampled star curve shows a sharp April launch, then steadier accumulation through May and early June. It crossed 13,000 stars on June 8 and reached about 16,000 by June 14, 2026. That is attention, not proof of retention, but it explains why Tolaria landed in the weekly trending list despite still carrying alpha-release caveats.

For another local-first knowledge tool aimed at AI research workflows, see Open Notebook. If your priority is converting files into Markdown before they enter a vault, MarkItDown is the practical companion. For codebase graphs rather than personal knowledge graphs, Understand Anything attacks the same “make a corpus navigable” problem from the developer side.

FAQ

What is Tolaria? Tolaria is a desktop app for managing Markdown knowledge bases. It stores notes as local files, uses YAML frontmatter for structure, and can layer Git history and AI-agent workflows on top.

How do I install Tolaria on macOS? The documented macOS command is brew install --cask tolaria. You can also download builds from the Tolaria download page or GitHub releases.

Does Tolaria work with existing Obsidian-style vaults? The docs say you can open an existing folder of plain Markdown files, including an Obsidian-style vault. Tolaria scans .md files and uses frontmatter when present.

Is Tolaria production-ready on Windows and Linux? It publishes builds for both, but its own docs call Windows and Linux support early. Recent issues mention Windows wikilinks, Linux image paths, IME input, and AI tooling edge cases, so test your workflow before relying on it.

Is Tolaria an AI app or a notes app? It is a notes and knowledge-base app first. The AI angle is that a local Markdown vault with conventions, Git, and agent setup paths gives Claude Code, Codex CLI, or Gemini CLI a clearer corpus to work with.

What license does Tolaria use? The repository is AGPL-3.0-or-later. That is friendly to users who want open software, but companies modifying or distributing it should review the license before adopting it.