Documentation Index
Fetch the complete documentation index at: https://docs.scoutvoice.ai/llms.txt
Use this file to discover all available pages before exploring further.
How audio works
Your audio goes straight from your microphone to the speech engine — it never passes through Scout’s servers.- Your microphone captures audio
- Scout sends the audio stream over an encrypted connection directly to the speech engine
- The speech engine transcribes in real time and sends text back to Scout
- Scout processes the text on your machine (voice commands, dictionary, quick phrases, app styles) and injects it at your cursor
What Scout’s server handles
- Authentication — Verifies your identity and subscription status.
- Temporary tokens — Short-lived tokens that let your app connect directly to the speech engine. Your top dictionary entries are included as recognition hints.
- Usage tracking — Session metadata (word count, duration) for billing. No audio or transcript content.
- Settings sync — Dictionary, quick phrases, and app style preferences.
- Semantic correction (Pro only) — For Pro users, transcript text may be sent to Scout’s server for real-time semantic correction (fixing misheard words). When context awareness is on, the request may also include a short excerpt of text immediately before the cursor or field tail, plus the active app label and window title. This data is used only to process the correction and return the result; it is not stored beyond the request lifecycle. Free users do not use this feature; their transcripts are processed entirely on-device.
- Magic Edit — When you use Magic Edit, your selected text and instruction are sent through Scout’s server to the AI model. This only happens when you explicitly activate Magic Edit.
Context awareness
When context awareness is enabled (the default), Scout reads text from the focused text field, selected text, nearby content in the focused app, and the active window title. When the app exposes true caret context, Scout may also read text immediately before and after the cursor. Key terms are extracted locally and sent to the speech engine as recognition hints.- Only the focused window is read — not other windows or background apps
- No screenshots are taken
- Raw active-window text is not stored on Scout’s servers
- If Pro semantic correction runs, Scout may send a short preceding-text excerpt, app label, and window title as part of that one correction request
- You can disable this in Settings → General
- Coverage varies by app and platform because different apps expose different accessibility information
What Scout stores
| Data | Where |
|---|---|
| Account info, subscription status | Server |
| Usage metrics (word count, duration) | Server |
| Dictionary, quick phrases, app styles | Server |
| Hotkey preferences, activity history | Local device |
| Context audit logs (dev builds only) | Local device |
What Scout does NOT store
| Data | Why not |
|---|---|
| Audio recordings | Streams directly to the speech engine, never sent to Scout |
| Transcripts | Not stored on Scout’s servers. Pro users’ text passes through for real-time correction but is not persisted. Free users’ text stays on-device. |
| Screen content | Only reads the focused window when context awareness is on; not stored on Scout’s servers |
| Keystrokes | Only intercepts Scout’s own hotkeys |
Security
- All network traffic is encrypted in transit
- Passwords are hashed using industry-standard algorithms
- Authentication uses short-lived tokens
Account deletion
Contact support@scoutvoice.ai to delete your account. This removes your account info, preferences, usage history, and cancels your subscription immediately.Your rights
- Access — View your data in the app
- Export — Contact support@scoutvoice.ai
- Deletion — Delete your account at any time
- Opt out — Disable context awareness in Settings