Create a new AI voice agent. Requires API key authentication via Authorization: Bearer or X-API-Key header.
Create a new AI agent.
Only name and system_prompt are required. All other fields default to the same values as the database defaults so callers can omit them.
Agent display name
1 - 100System prompt for the LLM
Optional description of the agent
LLM provider (e.g. 'openai', 'anthropic')
Model identifier
LLM sampling temperature
0 <= x <= 2Max tokens per LLM response
1 <= x <= 8192Opening message spoken by the agent
Who speaks first: 'agent_speaks_first' or 'user_speaks_first'
TTS provider
Provider-specific voice identifier
Voice model identifier
0 <= x <= 10 <= x <= 10.5 <= x <= 2STT provider
STT model identifier
BCP-47 language code
Preset background sound name
Custom background audio URL
Maximum call duration in seconds
10 <= x <= 43200Silence timeout in seconds
5 <= x <= 300Allow agent to end the call
Enable voicemail detection
Message to leave on voicemail
Webhook URL for server-side events
Tool IDs to attach to the agent
Enable retrieval-augmented generation
RAG configuration object
Enable HIPAA compliance mode
Record audio for this agent
Enable call logging
Save call transcripts
Successful Response
Full agent object returned by the API.