AI Agents
Agents are the core building blocks of the platform. They combine LLM intelligence, voice synthesis, and tools to conduct conversations.
System Prompt
The fundamental instructions for the agent. This defines the persona, tone, constraints, and knowledge. You can use variables like {{customer_name}} for dynamic injection.
Voice Settings
Configure the synthesizer provider (Cartesia, ElevenLabs, OpenAI), voice ID, speed, and stability. Different voices can be mapped to different emotions or agent states.
First Message
Define what the agent says immediately when the call connects. A clear opening sets the context for the user and reduces latency perception.
Configuration Options
| Parameter | Type | Description |
|---|---|---|
| name | string | Friendly name for the agent configuration. |
| model | enum | LLM Model ID (e.g., gpt-4o, gemini-2.0-flash). |
| transcriber | object | STT settings, including language and provider (Deepgram, AssemblyAI). |
| max_duration | number | Maximum call duration in seconds (hard limit). |