Vapi pricing in 2026, what a minute of voice agent really costs

Vapi costs $0.05 a minute in hosting fees, plus the speech to text, model, voice and telephony providers at their own list prices, according to Vapi’s published prices on October 19, 2026. Vapi’s calculator puts 1,000 minutes on Deepgram, OpenAI and ElevenLabs at $82 to $129 a month, so $0.08 to $0.13 a minute all in. You start for free with $5 of credits. The paid Success Packages add support, concurrency and data controls, from $29 a month.

Vapi is a hosted platform for voice AI agents. You configure an assistant with a transcriber, a model and a voice, then Vapi runs the conversation on its servers, over a phone line or from its web and mobile SDKs. We split the Vapi bill into lines below, work out one minute of conversation and the monthly bill at 1,000, 10,000 and 100,000 minutes, then price the same conversation on your own server.

Vapi pricing in short

A typical web agent comes to about $0.10 a minute on Vapi in October 2026, before the monthly fees for concurrency, support and compliance.

What the $0.05 hosting fee pays for

The hosting fee pays for the conversation loop Vapi runs on its servers, with turn-taking and interruptions. Every plan also includes unlimited agents, call logs, monitoring and simulated test calls. Vapi bills the same fee whatever the stack, for every minute the call runs, silences and the user’s turns included.

Everything that turns speech into text, text into an answer and the answer into a voice comes on top. Vapi’s pricing FAQ lists four components in the per-minute cost: “speech-to-text, the language model, text-to-speech, and transport. Each is billed at the provider’s listed price with no markup. Vapi’s $0.05/min hosting fee is added on top.”

You can also plug your own keys into Vapi for each provider. Vapi’s documentation on provider keys states that once a key is validated, “you won’t be charged when using that provider through Vapi”. The provider then bills you directly, at the same list price or at whatever rate your contract gives you. The hosting fee stays, since it pays for Vapi’s own work.

Transcription costs more on Vapi than the provider’s price per minute suggests. Vapi’s documentation on how it calculates cost states that Vapi “transcribes audio from both the caller and the assistant”, so its estimate counts two audio channels. Its calculator puts OpenAI transcription at $0.0118 to $0.0122 a minute of call, about twice the $0.006 per minute of audio that OpenAI lists for gpt-4o-transcribe.

A minute of conversation, line by line

Take a voice agent inside a web app. It runs on GPT-5.6 Terra, speaks with ElevenLabs Flash and transcribes with gpt-4o-transcribe, the same stack we used to price a minute on Retell AI, so the two bills compare line for line. The call starts from the browser, so the transport is free. A typical minute holds 25 seconds of user speech, 30 seconds of agent speech and four turns. The model reads a 1,500 token system prompt and the whole history again on every turn, so it takes about 10,000 input tokens and returns 100 output tokens a minute.

The model and voice prices come from OpenAI’s API prices and ElevenLabs’ API prices, both read on October 19, 2026. ElevenLabs counts about 1,000 characters per minute of audio, so 30 seconds of agent speech is about 500 characters. Vapi’s own estimate assumes five model requests per turn and 150 output tokens a minute, so its calculator gives a higher model cost than this example.

The hosting fee makes up half of this minute, a share that grows as the providers get cheaper. With a cheap model and a cheap voice, the $0.05 can make up three quarters of the bill.

Plans and add-ons that change the bill

Vapi replaced its Build and Scale plans with Success Packages in September 2026, a change the usagepricing.com tracker first recorded on September 20. Vapi’s documentation on billing and Success Packages states that the packages apply to new customers, while accounts on the old pricing keep their terms. Every package keeps the same usage billing, $0.05 a minute plus the providers. What a package adds is support, concurrency, data retention and compliance.

Three add-ons are available with any package, and without one:

• Each concurrent call line past the package default costs $10 a month.
• Each extra organization, a separate workspace with its own agents, numbers and keys, costs $20 a month.
• HIPAA costs $2,000 a month per organization, with a data processing agreement and a business associate agreement.

Three changes from the old plans matter for a budget. Zero data retention used to be a $1,000 monthly add-on and now comes with the $29 Core package. The free credits went from $10 to $5. A new account without a package gets 4 concurrent calls, where the old Build plan included 10.

Pro bills 10% of your hosting fees, so the $999 minimum applies until those fees pass $9,990 a month, about 200,000 minutes. Pro is a flat $999 a month below that volume, for the uptime commitment, the Slack support and the 30 lines.

Vapi gives you one free US phone number, five with Core and ten with Pro. Free Vapi numbers only take inbound calls. Outbound calls and campaigns need a number imported from Twilio, Vonage or Telnyx, which charge $0.0055 to $0.014 a minute of transport.

What Vapi costs at 1,000, 10,000 and 100,000 minutes a month

Multiply the $0.103 minute by the volume, then add the concurrent call lines you need. The number of lines depends on when people call. This example spreads the calls over 22 working days of 8 hours, with three times the average load in the busiest hour, so the peak reaches one call at once at 1,000 minutes, about 3 at 10,000 and about 29 at 100,000.

Core with 19 extra lines costs less at 100,000 minutes than 25 extra lines without a package ($250) or Pro’s $999 minimum. A team that wants the 99% uptime commitment would pick Pro, for a total of $11,299 a month. The hosting fee alone comes to $5,000 a month at that volume.

The same stack costs $0.159 a minute on Retell AI before add-ons, since Retell resells the model and the voice at a flat price per minute where Vapi passes tokens and characters through. You can compare Retell AI’s price for each provider line by line.

Running the same pipeline on your own server

The alternative to a platform fee is to call speech to text, the model and text to speech from your own server, on your own keys. The hosting fee disappears and the provider bills stay. Transcription costs less too, since the pipeline sends audio to the transcriber only while the user speaks.

The provider fees come to about $44, $440 and $4,400 a month at 1,000, 10,000 and 100,000 minutes. The gap with Vapi is about $0.06 a minute: the $0.05 hosting fee, plus transcription of the whole call on both channels. Those figures leave out the server and the engineering time, which are now yours. The providers you choose then set the price of each minute. A speech to speech model such as GPT-Realtime-2.1 replaces transcription, model and voice with audio tokens, about $0.048 for a minute shared evenly between the user and the agent, as we computed when comparing the OpenAI Realtime API with a pipeline.

You then run on your own server what Vapi ran for you: the hosting, turn-taking and interruptions, transcripts, the call history and a dashboard for whoever reviews calls, plus phone numbers if you need them. Micdrop is an open source TypeScript library for real-time voice conversations with AI. It takes care of the turn-taking and the interruptions. @micdrop/client handles the microphone, voice activity detection and playback in the browser, while @micdrop/server runs the conversation in your Node server with the providers you choose. Micdrop is MIT licensed, so the library itself is free. This server runs the stack of the worked example:

import { MicdropServer } from '@micdrop/server'
import { OpenaiAgent, OpenaiSTT } from '@micdrop/openai'
import { ElevenLabsTTS } from '@micdrop/elevenlabs'
import { WebSocketServer } from 'ws'

const wss = new WebSocketServer({ port: 8081 })

wss.on('connection', (socket) => {
  new MicdropServer(socket, {
    firstMessage: 'Hi, what can I help you with?',
    agent: new OpenaiAgent({
      apiKey: process.env.OPENAI_API_KEY,
      model: 'gpt-5.6-terra',
      systemPrompt: 'You answer questions about our product.',
    }),
    stt: new OpenaiSTT({
      apiKey: process.env.OPENAI_API_KEY,
      model: 'gpt-4o-transcribe',
    }),
    tts: new ElevenLabsTTS({
      apiKey: process.env.ELEVENLABS_API_KEY,
      voiceId: process.env.ELEVENLABS_VOICE_ID,
      modelId: 'eleven_flash_v2_5',
    }),
  })
})

Vapi’s fallback plans switch to a backup voice, transcriber or model when a provider fails. Micdrop’s FallbackTTS and its STT and agent counterparts do the same on your own server.

You can run a first call in about five minutes. Wiring Micdrop into an existing application takes a few hours. Transcript storage, the call history, the dashboard, telephony and hosting take longer, since you build them yourself. Micdrop has a single maintainer and no support contract. Products like Raconte.ai, built by Micdrop’s maintainer, and Cibli run on it in production. If you are weighing the move, we compared Vapi and Micdrop on architecture and migration.

When Vapi is worth its fee

Vapi is worth its fee when the product is phone calls. It gives you a US number for free, imports numbers from Twilio, Vonage and Telnyx, and comes with SIP, outbound campaigns and call routing already built. Rebuilding that telephony yourself would cost more than the hosting fee for a long time.

Vapi is also worth it when you need commitments you cannot give yourself yet: a 99% uptime commitment on Pro, 99.9% on Premier, a named support team, and a HIPAA business associate agreement for $2,000 a month. Check the region first if your data has to stay in Europe, since Vapi offers its EU region only with Premier.

Vapi also makes sense at low volume, or before you know the feature will hold up. The gap with your own pipeline is about $59 at 1,000 minutes a month, less than a day of engineering. It reaches about $6,100 a month at 100,000 minutes, enough to pay for the hosting, the dashboard and the person who maintains them. The other hosted platforms differ widely on price per minute, telephony and EU hosting.

Frequently asked questions

Vapi’s price is simple once you split it: $0.05 a minute of hosting, the providers at list price, then a package for support and concurrency. When the product is phone calls, or when you need an uptime commitment this quarter, you pay for work Vapi has already done. A voice feature inside a web app at volume costs less than half the Vapi bill with the same providers called from your own server, though you build the hosting and the dashboard yourself. Micdrop runs the conversation inside your Node server. You can start a first call in five minutes with your own provider keys.