Choose the workflow type. Recording: greet → instruct → record → thank. Live: real-time dialogue. No TTS: show instructions only, record up to 3 min, then save.
Greet → Instruct → Record → Thank → Process audio
Real-time conversation (OpenRouter: recording recommended)
Show instructions only + record button. Record up to 3 min → save.
Fields to collect from the user before they see instructions. Key = internal name; Label = text shown on the public agent page.
Describe what this agent will do and its goals. Recommended: 200-500 words.
Translated once into all languages when you save. Use Translate to convert to preferred language; Speak to hear how it sounds.
Only this text is shown to the user. No greeting or thank-you. Record button allows up to 3 minutes, then save.
Define what data you want to extract from conversations. Each field will be collected during or after the conversation.
Describe who receives the call, who arranged it, and how the AI should sound.
Review your configuration. You can navigate to any step to make changes.
Name your agent and save. All required fields must be filled.
Template for data points storage. Saved to a text file named after the table name.
Your agent is ready! Share this link to start using it. The link can be shared with anyone who needs to interact with the agent.
Agent name