Voice Agent
Send a voice clip or a message, get a spoken reply back as MP3 with the transcript.
A pay-per-call voice receptionist. Send a recorded clip as audioBase64 (MP3, WAV, OGG/Opus, WebM, FLAC, or M4A; up to ~768 KB on a public call) or a typed message as text. It transcribes the audio, writes a short spoken-style reply with the LLM, and reads it aloud, returning the transcript, the reply text, and a playable MP3 in one response. Edit the system prompt with your business facts (hours, services, policies) so it answers from what you gave it instead of guessing. Example input: { "text": "Hi, are you open this Saturday?" } Example output: { transcript: "Hi, are you open this Saturday?", text: "Thanks for calling! I don't have this week's Saturday hours in front of me. Can I take your name and number so someone confirms today?", mimeType: "audio/mpeg", fileName: "reply.mp3", fileBase64: "..." }
Who pays: Businesses and calling agents that want a spoken answer: voice notes, phone-tree front ends, and voice assistants that pay per reply instead of per seat.
Each accepted caller request triggers the 5-step flow once. Ordinary services may expose a free preview; payment-enabled calls may settle through x402; company or otherwise unready services may be unavailable.
The template opens with this intended price loaded. Launch publishes the service and its current call state. An ordinary service may preview; only a separately payment-enabled service exposes x402 terms for paid USDC calls.
The wired flow uses built-in nodes only. Any external action shows the Connection it needs for live use; a Connection is a reviewed link to your own webhook or account, and nothing external fires in preview.
Open Voice Agent in the studio
The 5-step flow arrives wired, with its suggested price loaded, and ready for a builder preview. Make it yours, then launch with its public call state reported.
Open this template →