Vapi
Voice AI infrastructure for developersDeveloper-first voice AI infrastructure for building, testing, and deploying real-time phone and web voice agents at scale.
- Maker
- Vapi AI
- Founded
- 2023
- Pricing
- Free tier · from $0.05/min
- Last checked
- October 2026
Verdict
Developer-first voice AI infrastructure for building, testing, and deploying real-time phone and web voice agents at scale.
Score breakdown
vs ElevenLabs, the category leaderEngineering teams building high-volume voice AI products who need full pipeline control
Key features
- Multi-LLM support: OpenAI GPT-4.1/Realtime, Claude 4, Gemini, and custom models
- Speech-to-text via Deepgram, Whisper, and other providers
- Text-to-speech via ElevenLabs, Play.ht, and others
- Real-time interruption detection and backchanneling for natural conversations
- Tool calling and webhook routing for mid-call backend actions
- Squad multi-agent call routing and escalation
- Phone number provisioning (US/Canada primary, international via workarounds)
- WebSocket streaming and dashboard for live call monitoring
Strengths
- +Model-agnostic architecture supports OpenAI, Anthropic Claude 4, Gemini, and more
- +Advanced real-time voice features: interruption handling, backchanneling, noise filtering
- +Squad feature enables multi-agent call routing and handoffs
- +Proven scale — over 1 billion calls handled as of 2026
- +Clean, detailed API documentation and active developer community
Weaknesses
- −Requires developer expertise to configure; non-technical teams are effectively locked out
- −Latency is inconsistent — typically 550-800ms but can spike to 4-5 seconds under load
- −Headline $0.05/min fee is misleading; true all-in cost with STT, TTS, LLM, and telephony runs $0.13-$0.31+/min
- −Customer support widely reported as poor or non-existent at self-serve tier
Overview
Overview
Vapi is a developer-focused voice AI infrastructure platform founded in 2023 by Jordan Dearsley and Nikhil Gupta, who previously went through Y Combinator (W21) with a productivity startup. Vapi was born when they pivoted away from an AI therapist product after discovering that startups wanted the underlying low-latency voice infrastructure more than the application itself.
The platform handles the full orchestration layer of a voice AI call stack: real-time speech-to-text, LLM inference, text-to-speech, telephony, and call routing. Developers connect their preferred providers — Deepgram for transcription, ElevenLabs for voice synthesis, OpenAI or Anthropic for reasoning — and Vapi glues them together with sub-second latency targets and human-like conversational features such as interruption detection, backchanneling, and noise filtering.
By mid-2026, Vapi has processed over one billion calls, raised $75 million in funding from Kleiner Perkins, Bessemer Venture Partners, and Microsoft M12, and reached a $500 million valuation after winning deployments like Amazon Ring. It is firmly positioned as the infrastructure choice for engineering teams — not a no-code tool. Non-technical buyers should look elsewhere; Vapi’s value is in low-level configurability and multi-model flexibility, not ease of use.
Key Benefits
- Model flexibility: Swap LLMs, STT, and TTS providers independently, avoiding vendor lock-in.
- Production scale: 1 billion+ calls processed gives confidence in infrastructure reliability at volume.
- Advanced conversation features: Interruption handling and backchanneling produce noticeably more natural voice interactions than simpler platforms.
- Multi-agent routing: The Squad feature lets calls be handed between specialized agents mid-conversation.
Use Cases
- Inbound customer support automation — Deploy 24/7 voice agents that handle FAQs, order status, and appointment booking with CRM integration via webhooks.
- Outbound sales and appointment dialing — Run high-volume outbound campaigns with dynamic scripting and real-time lead qualification.
- Voice interface for SaaS products — Embed conversational voice capabilities directly into web or mobile apps via WebSocket API.
- Healthcare and telehealth intake — Automate patient intake calls, triage questions, and appointment reminders with HIPAA-compliant configurations.
Pricing
Checked October 2026 · USD- $5 in starter credits on signup
- Full access to core features and dashboard
- No credit card required to start
- Vapi orchestration and hosting fee only
- Bring your own STT, TTS, and LLM API keys
- 10 concurrent calls included
- Additional concurrent lines at $10/line/month
- Dedicated infrastructure and custom SLA
- Priority support and onboarding
- Annual contracts typically $40,000-$70,000+
- Custom concurrency and volume pricing