Vapi is a developer-focused voice AI platform that lets teams build, deploy, and scale voice agents by combining their own choice of speech-to-text, large language model, and text-to-speech providers. It handles the hard real-time telephony infrastructure: endpointing, interrupt detection, latency optimization, and call routing. Squads allow chaining multiple specialized agents in a single call. It targets developers and technical teams building inbound/outbound voice automation at scale. Key features: - Bring-your-own-provider: plug in any STT (Deepgram, Assembly), LLM (GPT-4, Claude, Gemini), and TTS (ElevenLabs, Azure, Play.ht) - Squads: chain multiple specialized agents with handoffs within a single call - Function calling: trigger external APIs (CRM, email, SMS) mid-call - Knowledge base (RAG) support: upload PDFs/TXTs for live document Q&A during calls - Endpointing and interrupt detection for natural conversation flow - Concurrent call scaling: up to 10 concurrent calls on self-serve; unlimited on Enterprise
Pay-as-you-go: $0.05/minute platform fee (total cost $0.30-0.33/minute including third-party STT/LLM/TTS providers). Enterprise: custom pricing with unlimited concurrency.