Keebye is push-to-talk voice dictation for macOS, built for developers who run several AI coding agents at once. Hold Right ⌘, say the thing, release — and the text lands in whichever app has focus. Claude Code, Cursor, ChatGPT, Slack, a code editor, or a terminal session running over SSH inside tmux. Tap the hotkey once instead of holding it for longer dictations, and press Esc to cancel. If Right ⌘ isn't your key, Fn or Right ⌥ work just as well.
The whole speech pipeline runs on your machine. Audio never leaves your Mac: the network is used only to download models, check for updates and confirm your plan, and there is no telemetry or analytics of any kind. Once a model is downloaded, dictation keeps working with the Wi-Fi off, which makes it usable on planes, trains and bad hotel connections.
Terminal insertion is the part most dictation tools get wrong. Keebye pastes with terminal-aware chunking, and an optional synthetic typing mode survives SSH sessions, tmux panes and non-QWERTY keyboard layouts. Speech that arrives messy lands clean: deterministic rules strip fillers, false starts and doubled words, and an optional polish pass runs a local LLM entirely on-device. Turn cleanup off and you get every word verbatim — no model deciding what you meant.
The default engine is English-tuned and fast; switch to the multilingual Canary engine for 25 languages, or use Apple's native engine in your system language. A custom dictionary teaches Keebye your product names and acronyms once so every transcript afterwards spells them correctly. History is stored locally as text only and auto-deletes after 30 days, and Keebye refuses to type into secure password fields. Free 14-day trial, no card required.
