// MAC APP
The capsule
The capsule is a small floating bar that appears at the bottom-center of your screen whenever a supported AI app is open. Drag a file onto it and Omna returns a sliced, PII-masked version attached straight to your chat — all before you hit Send.
The capsule only appears when a supported AI app (Claude Desktop, ChatGPT, Codex, Cursor, or VS Code) is the active window, so it stays out of your way the rest of the time.
Idle state
When no file has been dropped, the capsule is idle — it shows rotating "Token Tax" tips and does nothing when clicked. Switch to a non-AI app and it hides itself; return to a supported AI app and it reappears.
Question first (recommended)
Typing your question before dropping the file gives Omna the most context and produces the sharpest slice.
- Type your question in the AI chat box — for example, "Which California customers spent over $10K last quarter?" — but do not hit Send yet. It just needs to be sitting in the text field.
- Drag your file onto the capsule. The moment it lands, Omna reads your question via the macOS Accessibility API, computes a content fingerprint, and checks whether the file is already indexed.
- Omna slices the file through the fast pre-indexed path or the live first-time path, finds the rows that match your question, and masks any PII before the rows leave your device.
- The sliced file attaches to the chat — an attachment chip appears, e.g. "customers-slice.csv · 261 rows · 49K tokens."
- Check the result card, then hit Send. Your question and the sliced file go to the AI together.
File first
If you drop a file before typing a question, the capsule enters Pending state instead of slicing immediately:
- The capsule shows the file name and waits — there's no timeout.
- Type your question in the AI chat box.
- Press Ready on the capsule. Omna reads your question at that moment and slices, then attaches as above.
The result card
After every successful slice, a result card appears (auto-dismissing after 10 seconds, or close it with ×):
| Field | What it means |
|---|---|
| Scenario | Killed / Crushed / Redirected / All rows matched — how big the savings are versus the AI's context window |
| Token reduction | Percentage cut, e.g. "−77% tokens" |
| Cost comparison | Estimated API cost without Omna vs. with Omna |
| Row counts | Rows sent in the slice vs. total rows in the file |
For the full explanation of the four scenarios, see How slicing works.
A note for Cursor and VS Code
These editors have an AI chat panel in the sidebar. The capsule reads your question from whichever text field is focused when the file lands.