Create your first agent
An agent is the voice worker that handles each call. It combines the conversation instructions, voice, speech settings, dynamic variables, tools, knowledge base access, phone configuration, retry behavior, and webhook targets.
Build path
Create the agent
Open the dashboard, go to Agents, and create a new agent. Give it a clear name that matches the workflow it owns, such as Renewal outreach, Lead qualification, or Support intake.
Write the operating instructions
Define the agent role, objective, tone, required questions, escalation rules, and when the call should end. Keep the prompt direct and operational.
Add call parameters
Add variables your backend will pass at call time, such as customer_name, account_id, appointment_time, or ticket_id. customer_phone is required for outbound calls.
Choose voice and speech settings
Select the voice, language, transcription provider, and turn-taking behavior that match your audience and latency requirements.
Production agent checklist
The prompt tells the agent what to collect, what not to say, when to transfer, and how to close the call.
Every variable used by the prompt, tools, results, or webhook payload is supplied by the API or CSV.
Expected caller questions retrieve the correct source content before the agent goes live.
Tool failures have fallback language and do not expose raw system errors to callers.
Your backend verifies webhook signatures and deduplicates events.
Structured outputs map cleanly into your CRM, BI, support, or workflow systems.