How Tolaria Integrates AI and Local Agents With Your Notes

Updated

Tolaria splits AI use into two distinct modes: tool-backed editing through CLI coding agents (Claude Code, Codex, OpenCode, Pi, Gemini), which can write to your vault, and chat-only access through local or API model providers, which read note context but get no vault-write tools. Which one you pick depends on whether you want the AI to change files or just talk about them.

The two AI modes

Mode Who runs it Can write to your vault? Typical use
CLI coding agents Claude Code, Codex, OpenCode, Pi, Gemini Yes — tool-backed editing Restructuring notes, bulk edits, agent-driven workflows
Local / API model providers Local models or direct API providers No — chat over note context only Asking questions about your notes, drafting ideas you paste in yourself

The distinction matters because it is the app's main safety boundary. If you only want an AI to reason over what's in your notes, the chat mode keeps it read-only by design. If you want an agent to actually edit files, you opt into a CLI agent that has write tools.

How the agent mode works

Tolaria is built on plain Markdown files with YAML frontmatter — no database, no proprietary format. That architecture is what makes external agents viable: an agent editing your vault is just editing files on disk.

When you use a CLI coding agent, the flow is roughly:

  1. Input — you invoke an agent (for example Claude Code) against your vault.
  2. Action — the agent reads and writes Markdown files and frontmatter directly, using its own tooling.
  3. Expected result — changes land as normal file edits, which you can then review, commit, and push through Tolaria's integrated Git client.

Because every change is tracked in Git, agent edits are reviewable and reversible like any other commit. That is the practical reason to prefer this setup over an opaque AI that mutates a hidden store.

How the chat mode works

For local or API model providers, Tolaria provides chat over your note context without vault-write tools. You supply the model (local or via a direct provider), and the conversation is grounded in your notes but cannot modify them. This is the mode to use when you want analysis, summaries, or questions answered without any risk of the AI rewriting content.

Organizing AI features

The sidebar supports custom sections, so you can group AI-related views and entry points separately from your notes. This is a layout convenience rather than a functional limit — it lets you keep agent and chat access within reach without cluttering your main note structure.

What to check before relying on this

  • Agent availability — the CLI agents listed (Claude Code, Codex, OpenCode, Pi, Gemini) are external tools; you need them installed and configured, not just Tolaria.
  • Model provider setup — chat mode requires you to point Tolaria at a local or API model provider. The site does not specify which providers or whether any keys are bundled, so treat provider access as your own responsibility.
  • Pricing — the site states "Free forever" and "no account required," and lists "subscribe" among its signals. It does not document paid AI tiers, so don't assume AI provider costs are covered by Tolaria itself.

If your goal is AI that can safely reshape your notes, use a CLI agent and lean on Git to review the result. If your goal is AI that only reads and discusses, use the chat mode and keep your vault untouched.