SpeaklySpeakly

Agent Mode

Double-tap the shortcut to talk to AI about what's on your screen

Agent Mode opens a floating chat bar that already knows what you are looking at: the moment you double-tap, Speakly captures the text visible in the active window — the e-mail you are reading, the page, the conversation — plus anything you selected. You speak your question or instruction and the answer appears in the bar, using your own AI model and your own key.

It has to be enabled
Turn on Enable assistant under Settings → System. With that off, a double-tap does nothing more than a normal recording.
Where the screen text goes
The captured text is sent only to the AI model you configured for post-processing — the built-in local model or your own API key. It is never stored, and it never touches Speakly's servers.

How to use it

  1. Select text in any app (optional — the visible window text is captured either way)
  2. Double-tap the same shortcut you dictate with, less than half a second apart — two quick taps, not two holds
  3. The Ask Speakly… bar opens already recording, grounded on the text visible in the window and on your selection
  4. Speak your question or instruction
  5. Press the shortcut to stop speaking — the answer appears in the bar
  6. Press Esc to close

The selection is context, not a target

Agent Mode answers in the bar: it does not replace the text you selected. If you want the text rewritten in place, use a custom prompt with normal dictation instead — that result is pasted over your selection.

Why the first tap leaves no stray recording

The first tap starts an ordinary dictation, as always. When the second tap arrives inside the double-tap window, that first recording is discarded and Agent Mode takes over. There is nothing for you to clean up.

Agent Mode — Speakly