Usage

Usage

Start AgentWhisper from your applications menu. A microphone icon appears in the system tray — that's the app. It runs quietly until you press the hotkey.

Dictating

Two recording modes, switchable anytime (tray menu → Recording Mode):

Mode	How it works
Hold to talk (default)	Hold F12, speak, release.
Press to toggle	Press F12 to start, press again to stop.

While recording:

• the tray icon turns red
• a translucent panel appears near the bottom of the screen with green bars moving to your voice — your confirmation that the microphone is live

When you stop, the tray shows Transcribing… briefly, then the text is typed into the window you were working in and placed in your clipboard as backup. A small notification shows a preview of what was heard.

Notes:

• While AgentWhisper runs, F12 is reserved exclusively — other programs won't react to it. Ctrl+F12 / Alt+F12 keep working normally.
• Very short taps (under ~0.3 s) are ignored as accidental.
• Changed your mind? Press Esc while recording to discard it — nothing is transcribed or typed. Esc is only reserved while a recording runs; the rest of the time it works normally everywhere.
• Recordings are capped at 60 s so a stuck key can't record forever — change the cap in the tray menu → Recording Limit (30 seconds to 10 minutes).

The tray menu (right-click)

Item	Effect
status line	What to do in the current mode / current activity
Enabled	Pause/resume dictation without quitting
Auto-Type into active window	On: text is typed for you. Off: clipboard only (paste with Ctrl+V)
Notifications	The "Typed & copied" confirmations on/off
Start at login	AgentWhisper starts with your session
Recording Mode	Hold-to-talk vs. press-to-toggle
Recording Limit	Max length of one recording (30 s – 10 min)
Choose active window…	Pick a window from a grid; every dictation is typed there + submitted with Enter (see below)
Quit AgentWhisper	Stop the app (releases F12)

Dictating into one window

Made for talking to an AI agent in a terminal while you work elsewhere: choose Choose active window…, pick the agent's window from the grid that pops up, and from then on every dictation is typed into that window and sent with Enter — regardless of which window is focused. Details:

• Delivery happens in the background: the target window is not raised. It borrows the keyboard only for the moment of typing, then focus returns to the window you were working in. Avoid typing during that split second — those keys would land in the target window.
• The clipboard still gets a copy of every dictation.
• If the target window closes, AgentWhisper notifies you and goes back to normal typing.
• Click the same menu item (now Stop dictating into: …) to end it.
• The choice is session-only — it is not saved to config.toml.

The command line (optional)

Everything the tray does, plus inspection:

agentwhisper status        # full health report: engine, hotkey, desktop tools
agentwhisper toggle        # enable/disable dictation
agentwhisper mode hold     # or: toggle
agentwhisper limit 120     # max seconds per recording (30-600)
agentwhisper target choose # send every dictation to one window (or: clear)
agentwhisper cancel        # discard the recording in progress (same as Esc)
agentwhisper autostart on  # start with your session (or: off)
agentwhisper quit

agentwhisper status is the first thing to run whenever something seems off — every subsystem reports its state there.