The Napster Omniagent API supports five channels. Each channel is a different transport — how you connect. You can deploy the same agent to multiple channels simultaneously.
Available channels
| Channel | Description | Modality |
|---|---|---|
| WebRTC | Browser-based audio and video via the Web SDK. Embeds directly into any web page. | Audio + Video |
| WebSocket | Persistent connection for custom integrations. Supports audio or text-only. | Audio or Text |
| VoIP | Answer phone calls via Twilio, Telnyx, Vonage, or any VoIP provider. No SIP trunk required — just a webhook URL. | Voice |
| SIP | Bring your own SIP trunk for enterprise telephony integration. | Voice |
| Kiosk | In-person, walk-up conversations on a Napster Station device. | Audio + Video |
Text-only sessions
WebSocket sessions support a text modality — a typed chat with no voice and no avatar. Set modality: "text" when creating a WebSocket connection. See Text-Only WebSocket Sessions for details.
Typical response latency: ~300ms across all channels. The platform supports approximately 30 languages.
Human handoff via SIP — prompt guidance
If your SIP agent is configured to transfer calls to a human agent, the timing of the handoff is controlled by your agent's prompt. A common issue is the agent handing off immediately at the start of every call before it has spoken at all. This is caused by a handoff instruction that is too broad.
- Make the handoff condition specific: "Transfer to a human agent only if the caller explicitly asks to speak to a person, or if you cannot resolve their issue after two attempts."
- Avoid broad triggers like "transfer if the caller needs help" — this can fire on every call.
- Test your handoff prompt in the Playground before going live.
Concurrency limits
Each organisation can run up to 5 concurrent sessions per channel type, for a total of 20. See API Billing and Pricing for details on limits and how to request higher limits.