Overview
What does Vapi do?
Vapi is voice agent infrastructure for developers. Rather than a single opinionated stack, it orchestrates the pipeline — transcription, language model, speech synthesis — and lets you choose the provider at each layer, including ElevenLabs for the voice itself.
That modularity is the entire point. You can route a simple intent to a cheap fast model and escalate complex turns to a frontier one, swap transcription providers when one handles your accents better, and tune the latency budget layer by layer. Teams running voice at real volume care about this because the difference between a 600ms and a 1.2s response is the difference between a conversation and an interrogation.
It is API-first with a dashboard on top, not the other way round. Expect to write code. In exchange you get per-minute costs below the bundled platforms and control over every part of the call.
Features
Vapi features
Swappable pipeline
Choose your transcription, model and voice provider independently, and change any of them without a rebuild.
Low-latency orchestration
Sub-second response times through streaming across every stage of the pipeline.
Function calling
Agents call your APIs mid-conversation for lookups, bookings and record updates.
Squads
Hand a call between specialised agents mid-conversation rather than building one agent that does everything.
Full telephony support
Inbound, outbound, transfers and voicemail detection, with your own Twilio account or numbers via Vapi.
Call analytics and recordings
Transcripts, structured extraction and per-call cost breakdown by pipeline stage.
Integrations
Integrations and how to connect Vapi
What Vapi connects to, and the order to set it up in.
Create an assistant
Define it in the dashboard or via API. The API is the real interface; the dashboard is for inspection.
Choose your pipeline
Pick transcription, model and voice per layer. Start with the recommended defaults and optimise only once you have real call data.
Add functions
Define the API calls the agent can make mid-call, with strict schemas. Loose schemas cause hallucinated arguments.
Connect telephony
Bring your Twilio account or buy a number through Vapi. Bringing your own is usually cheaper at volume.
Tune latency against real calls
Measure end-to-end response time over actual telephony and optimise the slowest stage. This is where the platform earns its keep.
Fit
Who is Vapi for?
Developers building voice into a product
The intended user. Full pipeline control and an API-first design.
Teams running voice at high volume
Per-layer optimisation meaningfully reduces per-minute cost once minutes are real.
Agencies deploying for many clients
Programmatic provisioning makes many similar deployments manageable.
Non-technical teams
Poor fit. This is infrastructure — use ElevenLabs Agents or Retell instead.
Use cases
Best use cases for Vapi
Voice features inside a product
Where voice is a component of something larger rather than a standalone agent.
High-volume call operations
Where shaving cents per minute across thousands of minutes is material.
Complex multi-stage calls
Squads route a caller between specialised agents rather than overloading one prompt.
Latency-critical applications
Where per-stage tuning gets you under the threshold at which conversation feels natural.
Pricing
Vapi pricing
Per-minute platform fee plus provider costs at cost. Free tier: Yes — 60+ minutes included on the Build plan.
Build
Most common$0.05/minute
Pay-as-you-go. Providers are billed at cost, or nothing if you bring your own keys.
- 60+ minutes included
- 10 concurrent calls
- Models at cost ($0 with your own key)
- SMS/chat $0.005/msg
Scale
Custom
Committed volume on an annual contract.
- Fixed platform fee
- SOC 2 and HIPAA
- SSO and RBAC
- Dedicated account team
Compliance add-ons
$1,000–$2,000/month
Priced separately on both tiers — easy to miss when budgeting.
- HIPAA $2,000/month
- Zero Data Retention $1,000/month
Pricing last verified . Vendors change pricing frequently — always confirm on Vapi’s own pricing page before buying.
User sentiment
What other users say about Vapi
A summary of publicly available feedback about Vapi — not paid testimonials.
Developers rate Vapi the most flexible voice platform available, and per-layer provider choice is cited as the reason both for quality tuning and cost control. The consistent criticisms are that total cost is hard to predict because it spans several providers, and that the flexibility means more ways to build something subtly broken.
What users praise
- Swappable pipeline gives real control over quality and cost
- Genuinely low latency when properly tuned
- Squads handle complex call flows cleanly
- Cheaper per minute than bundled platforms at volume
Common complaints
- Total cost spans several bills and is hard to forecast
- More configuration surface means more ways to get it wrong
- Documentation assumes real developer context
- Occasional platform instability reported under load
We publish individual user reviews only once they have been collected and attributed to a real source. None have been collected for Vapi yet. Used it? Send us your review.
Sentiment summarised from: Vapi — documentation, Vapi — pricing.
Verdict
Is Vapi worth it?
Pros
- Most flexible voice platform available
- Per-layer cost and quality optimisation
- Strong telephony feature coverage
- Free credit sufficient for genuine testing
Cons
- Requires engineering to use well
- Costs span multiple providers
- More configuration means more failure modes
- Reports of instability at high load
Assessed from vendor documentation, pricing pages and public user sentiment — not a hands-on trial. How we assess agents.
FAQ
Vapi frequently asked questions
What does a Vapi call actually cost?
Roughly $0.05/minute platform fee, plus transcription, model and voice provider costs, plus telephony. Realistically $0.07–0.15/minute all in depending on your pipeline. The dashboard breaks it down per stage, which is how you optimise it.
Vapi or ElevenLabs Agents?
ElevenLabs for the best voice quality with the least work — the right default when the caller is a customer. Vapi when you want to choose each pipeline layer, tune latency, or drive per-minute cost down at volume.
Do I need my own Twilio account?
You can bring one or buy numbers through Vapi. Bringing your own usually works out cheaper once volume is meaningful.
What are Squads?
A way to hand a call between specialised agents mid-conversation — a triage agent transferring to a booking agent, say — instead of writing one prompt that tries to cover everything.
Compare