Layercode offers a hosted backend option that makes it easy to build and deploy voice agents without writing the agent backend yourself. In this mode, Layercode handles the backend logic for you: when a user speaks, Layercode sends their query to an LLM (Large Language Model) using a customizable prompt, then streams the AI-generated response back to the user as speech.

​
Features

  • Responses are generated using the best low-latency LLM available (currently Gemini Flash 2.0)
  • Conversation history is stored in Layercode’s cloud
  • You can customize the prompt to change the behavior of the voice agent
  • Integrate with your web or mobile frontend, or connect to inbound or outbound phone calls
  • Still have complete control over the transcription, turn taking and text-to-speech voice pipeline settings

This approach is ideal for quickly getting started with Layercode and is sufficient for simple voice AI applications.

If you need more advanced functionality or want complete control over how responses are generated, read more about how to Connect Your Backend.

We’re actively working to expand the hosted backend’s capabilities. Stay tuned for upcoming features, including MCP tools.

​
Integrating with Your Frontend

The Hosted Backend can be used to build web and mobile voice AI applications by integrating with one of our Frontend SDKs. Learn how to Build a Web Voice Agent. Note that even when using the Hosted Backend, you’ll still need to implement an Authorize Client Session endpoint to authenticate user sessions.

​
Integrating with Telephony

The Hosted Backend can be used to build voice AI applications that can be used over the phone. Learn how to Build a Phone Agent (coming soon).