Kivo is a quiet background utility for system-wide dictation and focused writing
assistance. Hold a key, speak anywhere, release to insert — then proofread, rewrite,
or summarize without leaving your flow.
Fn hold to dictateCtrl + Shift + Space writing toolsesc to dismiss
Live in any app — no window switching
Writing Tools✕
✓ProofreadFix grammar & spelling
✎RewriteImprove clarity
≡ConciseShorten, keep meaning
✦SummarizeLinks, video, text
The Flow Bar listens. Writing Tools polishes. Both vanish when you’re done.
No accounts
No telemetry
OS-native speech
Keys stay in your keychain
How it works
Three gestures. Zero window switching.
Kivo lives in the background and appears only the moment you need it — next to your cursor, under your selection, exactly where you’re working.
1
Hold to dictate
Hold Fn (Mac) or your chosen shortcut anywhere — email, docs, code, chat. The Flow Bar appears and listens. Release to insert.
2
Select to improve
Highlight any text and hit your Writing Tools shortcut. Proofread, rewrite, shorten, or change tone — then Replace or Copy.
3
Summarize anything
Paste a link to a public article or YouTube video — or paste the text itself — and get a short overview in seconds.
Features
Everything you need. Nothing watching you.
Native speech recognition, careful text insertion, and one optional AI step — only ever when you ask for it.
System-wide dictation
Speak in any application. Kivo uses your operating system’s own speech engine — on-device where the OS supports it — and inserts the transcript at your cursor. Optional AI cleanup fixes punctuation and filler without changing your meaning.
Writing Tools
A compact popup for selected text. Eight focused actions, custom instructions, and Quick chat when nothing is selected.
Proofread
Rewrite
Concise
Friendly
Professional
Key Points
Summarize links & video
Drop in a public webpage or YouTube URL. Kivo retrieves it through Gemini’s URL context and returns a tight summary — with its source attached.
Quick chat
Nothing selected? The same popup becomes a minimal ask box. Ask anything; the answer stays put until you copy it.
Private by construction
Microphone audio never leaves the OS speech engine. Text reaches Gemini only for the action you invoke, with retention off. API keys live in Keychain / Credential Manager.
Privacy
Your words take the shortest possible trip.
No hosted backend, no provider abstraction, no analytics. The table shows exactly where each kind of data goes.
Microphone audio → OS speech engine only. Never sent to Gemini.
Selected text → Gemini only when you run an action, with store: false.
API keys are stored by the OS and never returned to the webview.
No logs of transcripts, selections, clipboard, or credentials.
Data
Destination
Voice audio
OS only macOS SpeechAnalyzer · WinRT speech
Dictated transcript
Local insert Optional Gemini cleanup if you enable it
Writing action text
Gemini, on invoke Direct HTTPS, retention off
API key
OS vault Keychain · Credential Manager
Platforms
Native on Mac and Windows.
Real native integration — Accessibility on macOS, UI Automation on Windows — not a browser extension or overlay hack.
macOS 26+
Signed & notarized app + DMG. Direct distribution — the App Sandbox is intentionally off so system-wide Accessibility can work.
SpeechAnalyzer · DictationTranscriber · Keychain
Windows 11 24H2+
MSIX package with identity (required for native speech). UI Automation + Win32 input, Credential Manager for keys.
WinRT speech · UI Automation · MSIX
Docs
Learn it in five minutes. Hack on it in an afternoon.
The user guide covers setup, dictation, Writing Tools, and troubleshooting. The developer guide explains the codebase and the reasoning behind it.