Vapi review
- Made by
- Vapi, Inc., San Francisco
- Founders
- Jordan Dearsley and Nikhil Gupta
- Funding
- $72M total, including a $50M Series B led by Peak XV (May 2026)
- Platform fee
- $0.05 a minute, plus provider costs at no markup
- Free credit
- $5, with 4 concurrent calls
- Paid packages
- Core $29/month; Pro from $999/month; Premier custom
- HIPAA
- $2,000 a month add-on per organization
- Phone options
- Free US inbound number, Twilio, Vonage, Telnyx, SIP, web calls
Vapi is a developer platform for building AI voice agents, the software that answers and makes phone calls for a business. Instead of a fixed stack, you choose each piece: the speech-to-text engine that hears the caller, the language model that decides what to say, the voice that speaks, and the phone carrier. Vapi runs them together in real time and handles interruptions, tool calls and transfers. You build mainly through its API and dashboard, and you can test agents with simulated callers before launch.
Vapi charges a $0.05-a-minute hosting fee and passes each provider's cost through without markup. It says it has supported more than 1 billion calls and 1 million developers, and it raised a $50 million Series B in May 2026. It ranks #3 in our best AI voice agents, behind ElevenLabs Agents and Retell AI.
Vapi is worth it if you have engineers and want full control over every model, voice and carrier. No rival we tested gives more choice, and the pass-through pricing lets you tune cost call by call. It is a poor fit for teams that want one simple price or cheap HIPAA coverage: the bill has five parts, the free tier allows only 4 simultaneous calls, and HIPAA costs $2,000 a month. Non-developers will find Retell AI or Bland AI easier.
Score breakdown
Best for
- Developers who want to swap models, voices and carriers freely
- Teams testing several model setups side by side on cost and quality
- Products that embed voice agents in their own apps through an API
- Companies with their own SIP trunk or carrier contracts
Skip it if
- Small clinics that need HIPAA on a tight budget
- Non-technical teams that want a no-code builder
- Buyers who want one predictable per-minute rate
What we like
- Mix and match almost any speech, model, voice and phone provider
- Provider costs passed through with no markup
- Deep API, CLI, MCP server and testing tools (Evals, Simulations)
- Free Vapi phone number and free SIP and WebRTC transport
- Faster median reply than Retell in the independent phone test
Watch out for
- HIPAA costs $2,000 a month on top of usage
- Five-part bill is hard to forecast
- Only 4 concurrent calls without a paid package or add-on
- Visual Workflows builder retired in August 2026
- Trustpilot reviewers report slow support and reliability issues
Key features, explained
Assistants and Squads
An assistant is one agent with one system prompt, tools and structured outputs, which suits booking or support lines. A Squad links several assistants that hand the caller to each other with the conversation context, for example a triage agent passing to a billing agent. Vapi retired its visual Workflows builder on 18 August 2026 and now recommends Assistants for simple agents and Squads for complex ones.
Bring any provider
Vapi works with transcription engines such as Deepgram and Gladia, models from OpenAI, Anthropic and Google, and voices such as ElevenLabs. When Vapi's own keys are used, it bills each provider at cost. If you add your own API key for a provider, Vapi stops charging for it and the provider bills you directly, which helps if you already have volume discounts.
Phone numbers and telephony
Every account gets one free Vapi phone number, but free numbers use US area codes and take inbound calls only. For outbound calls or other countries, you import numbers from Twilio, Vonage or Telnyx, or connect a SIP trunk. Vapi's own telephony, SIP, WebSockets and WebRTC browser calls carry no transport fee. Twilio and other carriers bill you separately, for example $0.008 a minute for Twilio inbound.
Tools and developer access
Function tools let an assistant call your APIs during a call, for example to check an order or book a slot. Everything in the dashboard is also available through the API, and Vapi offers a command-line tool (CLI) and an MCP server so coding assistants can read its docs. Vapi says its average latency is under 500 ms, though the independent test measured more.
Testing: Evals, Test Suites and Simulations
Vapi has three testing tools. Evals run mock conversations and check replies with exact matches, patterns or an AI judge. Test Suites have an AI tester call or chat with your agent from a script. Simulations use AI callers that follow instructions, then score the outcome, and you can run many in one batch. Together they let you catch regressions before a prompt change reaches real callers.
Compliance and data controls
Vapi says it is SOC 2, HIPAA and PCI compliant. Raw call data is kept for 14 days without a package, 30 days on Core and 180 days on Pro. Zero data retention (ZDR), where Vapi does not store call content, comes with any paid package. HIPAA-eligible handling, with a standard DPA and BAA, is a $2,000-a-month add-on per organization. Pro adds basic role-based access; SSO needs Premier.
Where Vapi ranks
Our scores come from our category rankings, where each tool is scored 0–10 on weighted criteria against its rivals.
| Ranking | Listed as | Position | Score |
|---|---|---|---|
| Best AI Voice Agents | Vapi | #3 of 9 | 7.8 |
What users say about Vapi
Opinions on Vapi split sharply by who is writing. On Product Hunt, where most reviewers are developers, it scores 4.9 out of 5 from 25 reviews. They praise how fast they could ship a working voice agent, the choice of models and voices, clear API documentation and quick product updates. On Trustpilot it scores 2.4 out of 5 from only 15 reviews, mostly from paying businesses with problems. They describe pauses of several seconds between turns, dashboard settings that did not save, waits of up to two days for a human reply, provider fees they did not expect and trouble removing card details. Even happy developers say production agents need better observability to trace failures. The small review counts mean neither score is conclusive.
Users praise
- Fast way for developers to ship voice agents
- Wide choice of models, voices and carriers
- Good API documentation
- Rapid product improvements
Users complain about
- Multi-second pauses on some calls
- Slow human support
- Provider costs on top of the advertised fee
- Dashboard edits not saving
- Weak tools for tracing failures in production
| Review site | Rating | Reviews | Checked |
|---|---|---|---|
| Product Hunt | 4.9/5 | 25 | 2026-09 |
| Trustpilot | 2.4/5 | 15 | 2026-09 |
Ratings are third-party aggregates and change over time. They are context, not part of our score.
Vapi charges a $0.05-a-minute hosting fee plus the cost of each provider you use, and most agents land around $0.08 to $0.14 a minute all-in. The provider costs are passed through at list price: Vapi's own estimate is $0.0095 to $0.0099 a minute for Deepgram transcription, $0.0077 to $0.0452 for an OpenAI model and $0.0146 to $0.0238 for an ElevenLabs voice. Vapi numbers, SIP and web calls have no transport fee; Twilio adds $0.008 to $0.014 a minute. New accounts get $5 of free credit.
Usage is all you need to start. Optional success packages add call lines, support and data retention: Core costs $29 a month (10 lines), Pro costs 10% of your hosting fees with a $999 monthly minimum (30 lines), and Premier is custom. HIPAA is a separate $2,000-a-month add-on, which dominates the bill for small healthcare users.
| Plan | Price |
|---|---|
| No Success Package (usage only) | $0 a month; $0.05/min hosting plus provider costs |
| Core | $29/month plus usage |
| Pro | 10% of Vapi hosting fees, minimum $999/month, plus usage |
| Premier | Custom |
| Add-ons (any package) | HIPAA $2,000/month; extra call line $10/month; extra organization $20/month |
Full Vapi pricing breakdown, hidden costs and money-saving tips
How the Vapi bill works
Every minute an agent runs, Vapi charges its $0.05 hosting fee plus four provider costs at list price:
- Transcriber (speech-to-text), for example Deepgram at $0.0095 to $0.0099 a minute.
- Model, for example OpenAI at $0.0077 to $0.0452 a minute by Vapi's estimate.
- Voice, for example ElevenLabs at $0.0146 to $0.0238 a minute.
- Transport: free for Vapi numbers, SIP and web calls; Twilio charges $0.008 inbound and $0.014 outbound.
On top of usage you can add a success package (Core, Pro or Premier) for support, SLAs, more call lines and longer data retention, and add-ons such as HIPAA. Worked examples are on our Vapi pricing page.
Speed: what the independent test found
Openbenchmarks timed the pause between a caller finishing and the agent starting to reply, using recordings of real phone calls and each platform's default setup for new accounts. Vapi's median was 1,558 ms (95th percentile 2,008 ms). That put it ahead of Retell (1,740 ms) but behind ElevenLabs (1,424 ms) and Telnyx (1,296 ms).
Because you choose every provider, your own result can be better or worse than this. A fast transcriber, a small model and a short prompt all cut delay. Vapi's homepage claims under 500 ms on average; measure on your own calls.
What changed in 2026
- May: Vapi raised a $50 million Series B led by Peak XV, bringing total funding to $72 million, and said it had passed 1 billion calls.
- 18 August: the visual Workflows builder was retired and existing workflows stopped running. Vapi points users to Squads instead.
- Pricing: usage is now separate from optional success packages (Core $29 a month, Pro from $999, Premier custom), with add-ons for extra call lines, organizations and HIPAA.
Who should skip Vapi
Skip Vapi if nobody on your team writes code, if you need HIPAA for a small practice, or if you want one fixed rate per minute. Retell AI includes a free BAA and a visual flow builder, Bland AI charges one all-in rate, and ElevenLabs Agents was faster in the independent test. See our Vapi alternatives for the full list.
- Pick a Squad over one giant prompt for complex calls. Smaller prompts per assistant keep model costs and reply delays down.
- Add your own API keys for providers where you already have discounts. Vapi then stops billing that provider and you pay the provider directly.
- Run a Simulation batch after every prompt change. It is cheaper to catch a broken booking flow with AI callers than with customers.
- Use the free Vapi number only for US inbound testing. It cannot make outbound calls, so import a Twilio, Vonage or Telnyx number for campaigns.
- If you still had agents on Workflows, rebuild them as Squads now. Workflows stopped running on 19 August 2026.
Vapi alternatives
Retell AI
Best for regulated businesses that want compliance and a visual flow builder
ElevenLabs Agents
Best for teams that want the most natural voice with simple pricing
Bland AI
Best for high-volume calling with one predictable rate
Telnyx Voice AI Agents
Best for cost-focused teams that want the phone carrier and AI together
Jargon explained
- Hosting fee
- Vapi's own charge for running a call, $0.05 a minute. Speech, model, voice and phone costs are added on top.
- Pass-through pricing
- The platform bills you what each outside provider charges, without adding its own markup.
- Speech-to-text (transcriber)
- The part of a voice agent that turns the caller's speech into text for the AI to read.
- Squad
- Vapi's way of linking several AI assistants in one call, each handling one stage and passing the caller on with the conversation so far.
- Zero data retention (ZDR)
- A setting where the vendor does not keep call recordings or transcripts beyond what is needed to process the call.
Frequently asked questions
What is Vapi used for?
Vapi is used to build AI voice agents for phone and web calls, such as customer support lines, appointment booking, lead qualification, outbound collections and candidate screening. It is aimed at developers who want to choose each speech, model and voice provider.
How much does Vapi cost?
Vapi charges $0.05 a minute plus the cost of the transcriber, model, voice and phone carrier. Vapi's own estimate for 1,000 minutes is $82 to $129 a month. Optional packages cost $29 a month (Core) or from $999 (Pro), and HIPAA is $2,000 a month. See our Vapi pricing guide.
Is Vapi HIPAA compliant?
Vapi offers HIPAA-eligible data handling with a standard DPA and BAA as a $2,000-a-month add-on per organization. Without it, you should not send patient information through Vapi.
Does Vapi need coding?
Mostly, yes. You can create assistants in the dashboard, but tools, integrations and Squads usually need a developer. Vapi retired its visual Workflows builder in August 2026.
Vapi or Retell AI: which should I choose?
Choose Vapi for maximum control over providers and a slightly lower platform fee. Choose Retell AI for a visual flow builder, 20 free call lines and a free HIPAA BAA. Retell ranks one place above Vapi in our best AI voice agents ranking.
How many calls can Vapi handle at once?
4 concurrent calls without a package, 10 on Core, 30 on Pro and custom limits on Premier. Extra lines cost $10 each per month on any package.
Sources
Every fact on this page comes from public information. Vendor figures are labelled as vendor claims.
- Vapi pricing (Vapi)
- Pricing and success packages (Vapi Docs)
- Creating free phone numbers (Vapi Docs)
- Provider keys (Vapi Docs)
- Workflows migration guide (Vapi Docs)
- Simulations quickstart (Vapi Docs)
- Evals quickstart (Vapi Docs)
- Vapi introduction (Vapi Docs)
- Vapi homepage (Vapi)
- Vapi raises $50m Series B led by Peak XV (The Next Web)
- Vapi voice agent latency benchmark (2026) (Openbenchmarks)
- Voice agent latency comparison (2026) (Openbenchmarks)
- Vapi reviews (Trustpilot)
- Vapi reviews (Product Hunt)
- Retell AI pricing (Retell AI)
- ElevenLabs API and agents pricing (ElevenLabs)
- Bland AI pricing (Bland AI)
- Deepgram pricing (Deepgram)
- Telnyx Voice AI pricing (Telnyx)
- LiveKit pricing (LiveKit)
- OpenAI API pricing (OpenAI)