Katexs
FAQ

Resources

FAQ

Short answers to the questions developers ask most.

How fast does a voice agent respond?

Each turn targets under 200 ms for transcription, under 300 ms to the model's first token and under 150 ms to the first audio. Real-world response time depends on the providers, the model, the prompt length and any tools used in that turn.

Can I use my own OpenAI or Anthropic account?

Yes. Add your key in API keys. See Bring your own keys.

Do I pay for calls that don't connect?

No. Voice is charged on the real call duration from the end-of-call report.

What happens when I run out of credits?

New calls are refused. If a call takes the balance below zero, live agents are suspended until you top up. Turn on auto top-up to avoid this.

Can callers interrupt the agent?

Yes, by default. You control how easily in Conversation behavior.

Can one agent work on phone and chat?

Voice and chat agents are separate so each can be tuned for its channel. They can share knowledge content and, if you turn on shared memory, each other's conversation history.

Is there a REST API?

The REST API is in private preview. See the API reference.

Is there an SDK?

Not yet. Use HTTP directly. The widget script and the chat streaming API cover website chat today.

Where are my calls stored?

In your workspace, visible only to its members. See Data flow.

Last updated November 2026 · Found a problem? Email support@katexs.com