Voice & text-to-speech
Spoken replies with /tts, voice notes in, and running a Teammate hands-free from your phone.
Your Teammate can talk
Every Teammate can speak its replies. Send /tts on in any chat and answers come back as audio as well — no setup, no API key, no configuration. Voice output works out of the box on every agent server.
How the audio arrives depends on the channel: Telegram gets native voice notes — the kind you tap and play inline — while Slack and Discord get audio file attachments. On Discord the command is also registered as /voice, because Discord reserves /tts for itself; typing /tts as message text still works.
The /tts commands
| Command | What it does |
|---|---|
/tts on / /tts off |
Turns spoken replies on or off |
/tts chat on|off|default |
Overrides the setting for the current chat only |
/tts latest |
Speaks the most recent reply, once |
/tts audio <text> |
Sends a one-off audio message without turning TTS on |
/tts status |
Shows whether spoken replies are on |
/tts limit <chars> |
Caps how much of a long reply gets spoken |
/tts summary off |
Long replies are summarized before being spoken; this turns that off |
Voice in: just send a voice note
The other half of the loop: if your Teammate's model understands audio, you can send it a voice note and it gets transcribed automatically — the transcript is treated exactly as if you'd typed it, and your Teammate acts on what you said.
Whether this works depends on the model your Teammate is running: models with audio understanding handle voice notes natively; models without it won't pick them up. If your voice notes are being ignored, that's the first thing to check — switch to an audio-capable model with /model, or adjust the Teammate's default in Model configuration.
Hands-free, end to end
Put the two halves together and you can run a Teammate without touching a keyboard: /tts on once, then talk to it in voice notes and listen to its replies. This works best on Telegram, where both directions are native voice notes — practical for delegating work from your phone, on a walk, or anywhere typing is the bottleneck.
For long-running work it pairs well with the steering commands — a spoken "actually, focus on the pricing section first" steers a running task the same way a typed message would. See Slash commands.