Fn to dictate. Fn × 2 to ask, write, and act — what would you use it for?

I built SayKuku around two gestures that work together on your Mac:

Fn — Voice Input. Dictate at the cursor in the text field you're already using.

Fn × 2 — Voice Agent. Tell it what you want done: generate a reply, work with text, answer a question, or take a supported action.

The second gesture is the part I'd love your feedback on. You don't need to select text or open a separate chat window to ask a question or draft something. Here are some examples of requests:

1. Draft where you work

'Draft a reply saying Thursday morning works.' SayKuku generates text for the current field. You can then double-tap Fn again and say 'make it shorter' or 'more friendly' to refine its last output.

2. Understand the context you allow

'Summarize this' or 'What does this mean?' Voice Agent can use selected text or accessible text in the focused window, when that context source is enabled. The answer appears in a card you can read, copy, or insert. Context availability depends on what the current app exposes.

3. Edit existing text

Select a paragraph and say 'translate this to English,' 'turn this into a list,' or 'fix the grammar.' SayKuku works on that text and checks the original target before writing. If it can't write safely, the result stays available to copy.

4. Keep the conversation going

Double-tap Fn again and say 'explain the second point.' Recent conversation context lets you follow up without repeating the whole request.

5. Use supported Mac actions

'Open GitHub,' 'search the web for…,' or 'run my named Shortcut.' Voice Agent can open links, start a browser search, and run an Apple Shortcut after confirmation. These are specific supported actions; the Shortcut defines the automation.

I wanted the move from speaking words to asking for an outcome to be one extra tap of the same key. Fn handles everyday dictation; Fn × 2 adds an assistant to the workflow you're already in.

SayKuku is free, native to macOS 15+, and open source under Apache 2.0. Bring your own Qwen API key; cloud model usage is billed to your account. Audio and allowed context go to your selected Qwen region. History and Memory stay on your Mac, and you can control the context sources.

Product video:

Download and details: https://say.anikuku.com/en/

Source: https://github.com/UllrAI/SayKuku

Which Fn × 2 request would save you the most time, and in which Mac app? I'd especially like to hear what you'd want an assistant to do while you're already typing or reading.