Voice

Browser voice calls can cover the workspace overview or a particular conversation. The phone control in the workspace list starts an overview call; the one beside a chat composer starts a call scoped to that conversation. Navigating the app does not change an ongoing call's target.

Voice needs the host's voice configuration, a secure browser connection, and microphone permission. Remote use requires HTTPS. Provider credentials belong on the host. Captions and the call sheet's text input let you participate without continuous speech.

Choose Slower, Normal, or Faster speech pace before starting. This is a speaking preference, not an exact playback speed.

Use an OpenAI API key with access to the voice service and backend model required by your installed version. A coding-provider sign-in does not supply this key. Check existing configuration without displaying secrets:

farero config status
farero config --help

To save the key, pass it on standard input:

farero config set voice.openai-key --stdin

Supply the key through your secret manager's standard-output pipe into this command. Do not put the key in the command arguments or a chat message. On macOS, if you have copied the key from your secret manager, the complete command is:

pbpaste | farero config set voice.openai-key --stdin

On Windows PowerShell, the equivalent is:

Get-Clipboard | farero config set voice.openai-key --stdin

Clear the clipboard after saving. Restart the app to load the setting, then check the configuration again:

farero service restart
farero config status

The status should report the OpenAI key as configured without printing it. Open a browser call, allow microphone access, and ask for a short workspace summary. A connected call with an audible or captioned response is the success check. Browser calling does not require configuring SIP, Twilio, or a public telephone webhook.

Illustrative.

Draft a message asking this agent to explain the current diff without editing.
Show me the target conversation and exact message before sending it.

Reading context does not send a coding prompt. Sending a prompt or creating a workspace requires the displayed preview and confirmation. SIP and public telephone calling are deferred. For connection or permission problems, see Troubleshooting.