Vapi is the "voice agent stack as a service" answer for builders who don't want to assemble Twilio + Deepgram + ElevenLabs + an LLM provider themselves. The platform handles the full pipeline — call ingress, speech-to-text, LLM inference with tool use, text-to-speech, and call routing — and lets you swap any component (any LLM, any TTS provider) without changing your application code. Pricing is per-minute usage-based, which scales linearly with call volume. For teams building voice agents at scale, Vapi has become a leading infrastructure choice — competitors include Retell, Bland, and rolling your own. The trade-off is some lock-in to Vapi's orchestration; the win is shipping voice agents in days instead of weeks. For prototyping or low-volume use, a free trial credits balance is enough; production deployments quickly need paid tiers.
Vapi is a developer platform for building voice AI agents — phone calls, web voice widgets, and custom voice apps — combining LLMs, TTS providers (ElevenLabs, Cartesia, OpenAI), STT (Deepgram, Assembly), and telephony (Twilio) behind a single API. Used by companies deploying voice agents at scale.