Glad you asked.
Here’s what to know.
The practical details, before you say hello.
Can I test an agent in my browser?+
Yes. Open the studio, connect a Gemini or OpenAI API key in Providers, select an agent, and press Start conversation. Your browser will ask for microphone permission. Headphones help avoid feedback.
What is the difference between the two engines?+
Gemini and OpenAI both stream speech in and out. Choose the provider, model, and voice per agent.
Does it speak Hindi and Hinglish?+
Set an agent to English, Hindi, or Hinglish. Hinglish asks the model to follow the caller’s mix of Hindi and English.
Do I need my own API keys?+
Yes. Provider usage is billed directly by your selected providers. Keys saved through this installation are encrypted on the backend and never returned to the browser.
What latency should I expect?+
The engineering target is 300–500ms from speech end to audible response. Actual results depend on network, region, model, provider, and turn detection. The studio displays measurements from your own sessions; this target is not a benchmark claim.
Can I interrupt the agent?+
Yes. Enable natural interruptions in the agent settings. The voice engine listens for a new turn and FlowChime clears queued playback when you interrupt.
Are telephone calls supported?+
Connect a VoiceLink number using the WebSocket API page in the dashboard, then choose which agent answers in Call routing. The FlowChime API must be publicly reachable over HTTPS.
What gets saved?+
Agent configuration, account settings, session metadata, and measured latency persist in this installation. Transcripts are saved only when both the agent and workspace settings allow it. Microphone recordings are not stored by FlowChime.