Voice Agent API

The Voice Agent API provides a single endpoint for streaming audio input and receiving processed audio output, handling transcription, speaker role identification, disfluency tagging, and context‑aware prompting. It is designed to let developers focus on the surrounding application while the service manages the underlying voice AI pipeline, including accurate capture of fillers, repetitions, restarts, and informal speech patterns.

Target users include developers building customer‑support bots, AI companions, clinical workflow tools, language‑learning assistants, phone agents, and coaching or training platforms. The API supports detailed audio tagging such as verbatim keyterms, speaker roles, and non‑speech events, enabling nuanced conversational analysis and precise transcript generation for specialized domains.

What distinguishes the service is its emphasis on real‑time, high‑accuracy voice processing combined with an “invisible” infrastructure model: developers send audio streams and receive enriched audio or transcript data without needing to orchestrate separate transcription, tagging, or context‑management components. The offering is positioned as experimental and aimed at rapid prototyping of production‑ready voice agents.