Talk to your agents.
The voice control layer for Noml. Dictate prompts to your agents, drive the whole app out loud, and transcribe anywhere you type.
Coming soon
hatching
$ noml roar
press and hold to roar
Dictate promptsstraight into the composer
Drive the appwith voice commands
Transcribeanywhere you type
Command the whole app
Accept or reject changes, answer permission prompts, switch sessions, pick models and modes, open panels, spawn new agents. If Noml can do it, you can say it.
Hey Cera, keep listening
Hold a hotkey, or wake Cera by name. Conversation mode holds the mic between commands, and voice detection ends each take cleanly, so a review session stays hands-free.
Local and private
Recognition runs on a local Whisper model, offline. Your voice never leaves your machine, matching Noml's zero-telemetry stance.