Thought capture and review — a minimalist tool for quickly capturing thoughts, ideas, and tasks. Available on macOS, Linux, and Windows.
brew tap thawts/tap
brew install thawtsGrab the latest binary for your platform from GitHub Releases.
Run thawts to start the app. It lives in your system tray — no Dock icon. Register it to start automatically on login:
thawts --install| Action | Shortcut |
|---|---|
| Show/hide capture window | Ctrl+Alt+Space |
| Open review | Cmd+Option+R (macOS) |
| Hide window | Esc or click away (macOS/Linux) / Esc or hotkey (Windows) |
Run from any terminal:
thawts --tui| Key | Action |
|---|---|
Tab / Ctrl+R |
Switch to review |
Esc |
Review → capture; quit from capture |
Enter |
Save thought / confirm edit |
d d |
Delete (arm then confirm) |
e |
Edit selected thought |
/ |
Focus search |
thawts --install # register thawts to start automatically on login
thawts --uninstall # remove thawts from login items
thawts --tui # run the terminal UI instead of the GUI
thawts --version # print version and exit--install uses the appropriate mechanism for each platform:
| Platform | Mechanism |
|---|---|
| macOS | ~/Library/LaunchAgents/ (launchd) |
| Linux | systemd user service, or ~/.config/autostart/ (XDG) |
| Windows | HKCU\Software\Microsoft\Windows\CurrentVersion\Run (registry) |
Thawts includes a local AI model that runs entirely on your device. No internet connection is required after the first launch, and no data ever leaves your machine.
The model (all-MiniLM-L6-v2, ~22 MB) is embedded directly in the binary. The ONNX Runtime library (~30 MB) is extracted to your OS cache folder silently on first launch.
| Feature | How it works |
|---|---|
| Auto-tagging | Classifies thoughts into categories (todo, idea, question, calendar, reminder, finance) using semantic similarity |
| Sentiment tracking | Scores the emotional tone of each thought for the wellbeing trend view |
| Semantic search | Finds thoughts by meaning, not just keywords (embeddings stored locally in ~/.thawts/thawts.db) |
| Mishap detection | Flags accidental captures (passwords, code snippets, large pastes) |
AI is enabled on macOS (arm64) and Linux (amd64/arm64). Windows uses a regex-based fallback; full AI support is planned.
All data is stored locally at ~/.thawts/thawts.db (SQLite). Nothing leaves your machine.
Type / in the capture or search input to open the command palette:
| Command | Description |
|---|---|
/export json |
Export all thoughts to a JSON file |
/export csv |
Export all thoughts to a CSV file |
/import json |
Import thoughts from a JSON file (additive — keeps existing data) |
/import json restore |
Import from JSON and replace all existing data |
/import csv |
Import thoughts from a CSV file (additive) |
/import csv restore |
Import from CSV and replace all existing data |
/restart |
Restart the application |
/quit |
Quit the application |
Use arrow keys to navigate the palette, Tab to autocomplete, Enter to execute, Esc to dismiss.
The JSON format preserves all fields including tags, context metadata, and intents. The CSV format covers the core thought content and is useful for importing from external sources — only a content column is required; created_at, tags (pipe-separated), raw_content, hidden, window_title, app_name, and url are optional.
# Install prerequisites
brew install go node # or your platform's package manager
# Clone and build (stub AI — no downloads needed)
git clone https://github.com/thawts/thawts
cd thawts
npm ci --prefix frontend && npm run build --prefix frontend
go build -o thawts .
# Build with on-device AI (macOS / Linux)
bash scripts/download_ai_deps.sh # downloads model + native libs (~90 MB, one-time)
go build -tags with_onnx -o thawts .MIT