OWL is a desktop app that brings together the tools you use to think and work — notes, a task manager, PDF Q&A, time tracking, meetings, and reminders — with an AI assistant that runs entirely on your own computer.
Most "AI note apps" send your notes to someone else's servers. OWL doesn't. The language model runs locally (bundled, or your own Ollama), your notes are plain Markdown files on your disk, and sensitive notes can be encrypted behind a master passphrase. If you pull the network cable, OWL keeps working.
Who it's for: anyone who wants a capable, AI-powered workspace but isn't comfortable uploading their private notes, client work, or documents to the cloud.
- 🔒 Private by default — notes, documents, tasks, and AI chats stay on your device. No account, no sync-to-cloud, no telemetry.
- 🤖 A local AI that actually helps — ask questions about your own notes and PDFs and get answers with citations, powered by a model running on your machine.
- 📝 Notes you own — a folder tree of plain Markdown files you can read, back up, or edit with any other tool. Rich-text and raw-Markdown editing, autosave, full-text search.
- 📄 Chat with your PDFs — drop in a document; OWL extracts the text locally and lets you ask about it.
- ✅ Tasks & projects — projects, subtasks, due dates, blocked flags, drag-to-reorder, auto-archive on completion.
- ⏱️ Time tracking & My Day — per-task timers and a daily report of where your hours went, across tasks and meetings.
- 🔗 Jira integration — link tasks to issues, push worklogs, post comments (or prepare them to paste manually).
- 🔔 Reminders & notifications — due-soon/overdue nudges and long-running-timer alerts, as native desktop notifications.
- 🎨 Themeable — Soft Peach, Off-White, and Charcoal (dark) themes.
- 🔐 Encryption — protect the vault with a master passphrase; give individual notes their own password.
✈️ Fully offline — works with no internet connection.
This is the core promise, so here's exactly how it works:
- The language model runs as a local process on your computer — either a bundled engine (llama.cpp) or your own Ollama install. You pick the model; it's downloaded once and stored locally.
- "Ask about my notes" uses retrieval-augmented generation (RAG) entirely on-device: OWL searches your notes with a hybrid of keyword and semantic search, then asks the local model to answer using only the matching passages — and shows you which notes it used.
- The only network request OWL makes on its own is an optional check for app updates (which you can turn off in Settings). Your notes, documents, and conversations are never sent anywhere.
- Encrypted notes are never indexed or fed to the AI in plaintext.
| Area | What you get |
|---|---|
| Notes | Markdown folder/page tree, dual rich-text + Markdown editor, autosave, full-text search, import .md files, per-note or master-passphrase encryption, export to Markdown/PDF. |
| AI Assistant | One docked panel with four modes — Chat (free chat), Notes (cited Q&A over your vault), Documents (Q&A over PDFs), and Tasks (create tasks/subtasks/comments by asking). |
| Dashboard | In-progress tasks across projects, today's reminders, and colour-coded sticky notes. |
| My Day | Log meetings and see a daily time report (tasks by project + meetings) as proportional bars. |
| Documents | Import PDFs into folders, read the extracted text, and ask questions about them. |
| Tasks & Projects | Statuses, due dates, blocked reasons, drag-to-reorder, auto-archive, subtasks, comments. |
| Time tracking | Per-task timers with per-day records; log straight to Jira. |
| Costs | Per-project expenses with CSV export. |
| Jira | Per-project setup (API token, session cookie, or manual), worklog push, comments, ticket links. |
| System tray | Quick actions, live status, close-to-tray, graceful shutdown. |
Platform: OWL is currently distributed for macOS (Apple Silicon and Intel). It's built on Tauri and the codebase targets macOS first; Linux/Windows builds are possible from source but not officially packaged yet.
- Download the latest
OWL_x.y.z_<arch>.dmgfrom the Releases page. - Follow the installation guide from OWL Installation Guide page.
OWL is not code-signed with an Apple Developer certificate yet, so when you download the .dmg, macOS quarantines it and Gatekeeper shows a misleading "OWL is damaged and can't be opened" message. The app is not damaged — macOS is just refusing to run an unsigned, downloaded app.
To open it, run this once in Terminal after moving OWL to Applications:
xattr -dr com.apple.quarantine /Applications/OWL.appThen open OWL normally. (This is a one-time step; updates installed from within the app are not affected.)
- OWL downloads a small AI model on first use (you choose which in Settings → Local AI). The default is tiny and fast; larger models give better answers at the cost of RAM and disk.
- Prefer to use your own models? Install Ollama, then pick the Ollama provider in Settings — OWL will start and stop it for you.
- For "ask my notes" semantic search, download the embedding model in Settings → Notes Q&A and click Rebuild semantic index. (Keyword search works without it.)
OWL can check for new versions and update itself (downloads are verified before installing, and it never installs without asking). You control this in Settings → Updates:
- Check automatically (default) or only when you click.
- Download in the background, then prompt before installing — OWL never restarts on its own.
- Local-only data. Notes are Markdown files under your app-data folder; structured data (tasks, time, etc.) is a local SQLite database. Nothing is uploaded.
- Encryption. A master passphrase (Argon2id-derived key) unlocks the vault. Individual notes can be encrypted with AES-256-GCM, optionally under their own password. A forgotten passphrase can't be recovered — that's the point.
- No telemetry. OWL doesn't phone home. The only outbound request it initiates is the optional update check.
Tauri v2 · Rust · Vue 3 + TypeScript · Vite · Pinia · SQLite (FTS5 + sqlite-vec) · llama.cpp / Ollama · Tiptap + CodeMirror.
Is my data sent anywhere? No. Everything is local. The only request OWL makes on its own is an optional update check you can disable.
Do I need a GPU or lots of RAM? No GPU required. The default model is small and runs on modest machines; you can choose larger models if you have the RAM.
Can I use my own models? Yes — switch to the Ollama provider in Settings and use any model you've pulled.
Where are my notes stored?
As plain .md files in OWL's app-data vault, so you can back them up or read them with any editor. On macOS: ~/Library/Application Support/dev.owl.app/vault/.
Why does macOS say the app is damaged? It isn't — it's unsigned. See the install note above.
Is there a Windows/Linux version? Not packaged yet, but can be made avilable.