Vapi

What Vapi is

Vapi is a platform for building, deploying and improving voice agents. Its homepage pitch is "Voice agents for builders. Your agents, your rules." It sits closer to the developer end of the voice AI spectrum than a point-and-click tool, and the platform page sums up the split: "Full API access for developers. A powerful UI for everyone else." The company behind it is Vapi, Inc. The pages do not say where it is based.

The reader it is for is a developer or small technical team that wants control of the whole voice stack, plus business teams that want to configure and monitor agents in a dashboard. The platform page puts the trade this way: "All the control of code. None of the infrastructure overhead."

What you build and how

Voice agents on phone and web. The homepage lists customer support, lead qualification and appointment scheduling as demo scenarios, and you can ring a demo number or start a call in the browser. The pages describe voice only. They mention no text chat widget, so the chatbot-builders reader should treat Vapi as a voice product.

Prompt to production. The platform page promises to "deploy production-ready voice agents in minutes, then customize every interaction as you scale". You give the agent instructions and tools, then configure voice, conversation flow, telephony and integrations. The homepage says Vapi "handles the infrastructure so you can go from prompt to production fast".

Models and providers. You can "bring your own model, or choose from our managed list of 200+ models and integrations". You can also bring your own API keys for transcription, the language model or text-to-speech, or plug in self-hosted models. The platform is described as modular: "Swap models, voices, and logic independently, without vendor lock-in". Model fallbacks add "automatic failover and latency balancing across model providers".

Conversation quality. The platform page claims responses in under 500ms for real-time conversation, along with turn-taking, human-like voices and custom pronunciation. Telephony control is listed as custom SIP support, warm transfers, voicemail detection, phone number management and DTMF support.

Multi-agent setups. Multi-agent orchestration lets you "chain agents or use squads for complex workflows with intelligent handoffs and routing".

Testing and improvement. Every plan includes critical monitoring and testing and simulation, which lets you "simulate calls before they reach your customers". A/B experiments let you test variations of prompts, voices and flows. The homepage adds that you can "track what's working across every call, surface what isn't".

The API. The platform page says "Every capability is accessible via REST API. Your team owns the integration end to end." It links client and server SDKs, a GitHub presence and docs. The pages do not list specific CRMs or business systems, so integrations beyond models, voices and telephony carriers are a question for the company.

What the pages say about the product

Pricing. The pricing page has a minutes calculator and a clear structure. A usage package is called "No Success Package" and described as "Get started for free, and then pay only for the minutes that run". It is "$0 to start", pre-paid, with "$0.05/min Vapi hosting" and models passed through at cost. It includes $5 in free credits, 2 organizations, 1 included Vapi phone number, 4 concurrent calls, 14 days of raw data retention, the full model catalog and full platform access. Support is "peer to peer support via Discord community" and email support without an SLA. The Core package costs "$29 / month", pre-paid and auto-renewing, with models at cost, 2 organizations and 5 included Vapi phone numbers. A higher tier lists custom concurrency and data retention, priority support, a named support team, a 1 hour P0 response SLA, a 99.9% uptime SLA, custom RBAC, ZDR and SSO, and dedicated SIP.

The page's sample estimate for 1,000 minutes shows a $50 hosting fee, $10 for the Deepgram transcriber, $8 to $45 for the OpenAI model and $15 to $24 for ElevenLabs voice. Listed ranges are $0.0095 to $0.0099 per minute for transcription, $0.0077 to $0.0452 for the model and $0.0146 to $0.0238 for voice. Vapi telephony, SIP, WebSockets and Daily WebRTC are free. Twilio is $0.008 per minute inbound and $0.014 outbound, Vonage is $0.00814 and Telnyx is $0.0055. The page says transport is charged by the provider, not Vapi.

Add-ons apply to any package. HIPAA compliance, described as "HIPAA eligible data handling" with a standard Data Processing Agreement and Business Associate Agreement, is $2,000 per month. Extra concurrency is $10 per line per month, and extra organizations are $20 per organization per month. Every plan includes unlimited agents, models at provider cost with no markup, and "You own and operate your agents, models, and data".

Scale and compliance claims. The homepage reports 1 billion calls supported, 99.9% uptime for enterprise clients, 2.5M+ agents launched, 1M+ developers and under 500ms average latency. It lists "SOC 2, HIPAA, and PCI compliant" under enterprise capabilities, alongside SSO, OAuth, RBAC, AI guardrails and dedicated deployment support, "A dedicated engineer embedded in your team to get you live in a week".

Customers. The customers page names Amazon Ring, which "went with Vapi over 40 vendors", Kavak, UnityAI, Spring Venture Group, Instawork, Fleetworks and Ancile Services. Ring's VP of Software Development says "We went from zero to production in two weeks, and 100% of our inbound volume now runs through Vapi." UnityAI reports 5X revenue growth and 250k monthly calls. Spring Venture Group lifted Medicare conversions 25%, and the page says 4 engineers ramped to the volume of 50 call center agents in 1 week. Fleetworks runs 10K+ AI calls a day. These are company-selected stories.

Who it suits

Vapi suits developers, agencies and product teams that want to ship a voice agent and keep control of models, providers and data. The itemized pricing helps a team work out per-minute cost before it commits, and the free-to-start usage package lets a builder try the platform with $5 of credit. Enterprises get a defined path with SLAs, SSO and a compliance add-on.

It suits operations teams less when no engineer is involved. The pages talk about a UI "for everyone else" but make the API the center of the product. Per-minute costs are the sum of several providers, so the monthly total depends on the models and voices you choose rather than one flat number. It is also a voice product, not a website chat builder.