A Voice API is a programmable interface that lets developers initiate, receive, control, and automate voice calls and audio streams from software. Instead of building telephony infrastructure, developers make API calls to place or answer calls, stream real-time audio to and from speech models, play or record audio, and route conversations — usually over PSTN, SIP, or WebRTC. Popular voice APIs expose speech-to-text, text-to-speech, and call-control primitives so an AI agent can talk to callers programmatically.
Voice APIs are the backbone of AI-powered calling: voice AI platforms use them to build outbound campaigns, AI receptionists, and inbound call handling that connect large language models to real phone calls. Because they abstract away carriers and codecs, a developer can ship a production voice agent in days rather than months. Key things to compare are latency, audio quality, streaming support, concurrency, and telephony coverage.