docsyour bots
voice and calls
have a bot read its replies aloud, talk to it instead of typing, or call it and have a conversation.

bots can talk. you can listen to what ran overnight while you make breakfast, or call a bot and talk a problem through.
hear a reply
press the speaker on any reply and the bot reads it to you. you can also switch a bot to read every answer aloud as it arrives.
a reply is tidied up before it is spoken. the bot says the words and names the files, and skips the code and the punctuation that only make sense on a screen.
call a bot
choose call and it becomes a conversation. the bot hears you, tells you what it is doing while it works, and asks for approvals out loud.
while you are on a call:
- hold the talk key to speak. you choose the key in settings.
- Space interrupts the bot.
- Esc hangs up.
pick a voice
open a bot's settings and look for Voice & alerts. give each bot its own voice and a group chat stops sounding like one person.
voices come from one of these:
| voice | what it needs |
|---|---|
| built-in mac voices | nothing. they are part of macos. |
| elevenlabs | an elevenlabs api key |
| fish audio | a fish audio api key |
| grok (xai) | an xai api key |
| chatterbox | a chatterbox server you run yourself |
you paste a key once and every bot can use it. like every other key, it is stored on the machine your bots run on and is never shown again.
talk instead of typing
the microphone in the message box turns what you say into text.
- on a mac it listens on the device, with apple's own speech recognition. nothing is sent anywhere.
- on windows and linux it uses fish audio's speech-to-text, with the same fish audio key as the voices.
speech costs are billed by whichever voice service you chose, on your own key. the built-in mac voices and a local chatterbox server cost nothing.
stuck? write to [email protected] and a person answers.