Voice Input
Talk to your agents with speech-to-text in the composer.
AgentCraft supports voice input in the composer. Hit the mic button, speak your prompt, and get real-time transcription with auto-send on silence.
Using Voice Input
- Click the microphone button in the composer (next to the send button)
- Speak your prompt — text appears in real time as you talk
- Pause speaking — after a brief silence, the message auto-sends
You can also click the mic button again to stop recording manually before auto-send triggers.
How It Works
Voice input uses the Web Speech API built into modern browsers. Speech is transcribed locally by the browser — no audio is sent to AgentCraft servers.
Real-time transcription updates the composer text as you speak. When silence is detected for a configurable duration, the message is automatically submitted to the selected agent.
Browser Support
Voice input requires a browser that supports the Web Speech API:
| Browser | Support |
|---|---|
| Chrome / Chromium | Full support |
| Edge | Full support |
| Safari | Supported (macOS & iOS) |
| Firefox | Limited support |
Desktop & Mobile
Voice input works on both the desktop UI and the mobile PWA. On mobile, it's especially useful for quick replies and approvals while away from your keyboard.