BestOnline.ToolsBestOnline.Tools
Audio and voice

Vapi

Build a configurable voice-agent pipeline

Vapi is a developer platform for phone and web voice agents, combining speech recognition, language models, speech output and tools in a configurable call workflow.

voice-agent orchestrationPhone and web agentsDeveloper voice APIs

Paid plansAccount required

Product previewView original
Vapi's assistant publishing dialog compares a draft opening message with the current published configuration.
Vapi's assistant publishing dialog compares a draft opening message with the current published configuration.Vapi
Access
Paid access
Available on
Web, API
OVERVIEW

What is Vapi?

Vapi connects speech recognition, a language model, voice output and tools into an interactive application. Developers choose the providers and integrations for a phone or web experience.

An Assistant combines a prompt, tools and structured outputs for one workflow; Squads coordinate specialised assistants and handoffs. Completing a business action still requires a working connection to the relevant service.

Match the tool to the service contract

API Request tools call an existing HTTPS service directly. Function tools use Vapi’s tool-call protocol for server, asynchronous or supported browser-side execution. Choose the contract your service supports and validate the returned result before the assistant confirms a booking or update to a caller.

Separate release control from test coverage

Draft changes stay separate from live calls until publication. Restoring an assistant creates a current version immediately; restoring a tool loads a draft for later publication.

Chat simulations test logic without speech processing; Voice simulations exercise transcription, speech and turn-taking. Listen to the audio and verify real downstream actions separately from mocked tool results. Call monitoring supports continued review after release.

Budget the complete call configuration

Hosting sits alongside recognition, model, voice and telephony costs. Optional Success Packages add support and capacity without replacing usage charges. Estimate the full configured pipeline rather than treating one provider rate or support fee as the price of a complete call.

Best for

  • Choosing providers for phone and web voice agents
  • Connecting existing APIs or custom tool endpoints
  • Owning testing, rollout and call-cost monitoring

Limitations

The platform is a cloud service; a voice model or phone number alone does not complete the application integration. Hosting, model, voice, transcription and transport costs can be separate. Optional support packages do not include unlimited call usage.

WHAT YOU CAN DO

Vapi features

Configurable voice pipeline

Choose speech recognition, model and speech-output components for a conversational agent.

Phone and web integration

Build incoming/outgoing call experiences or embed voice calls in an application.

Tools and assistant handoffs

Connect APIs and use Assistants or Squads to organise a workflow.

Versions, simulations and monitoring

Review and improve a configured agent using the platform’s testing and call-observation tools.

Two business-action integration routes

API Request calls an existing HTTPS API; Function tools support Vapi-aware webhooks and other documented execution modes. Match the response contract to the tool type.

Drafts and two simulation modes

Publish a reviewed version for live calls. Use Chat simulations for logic and Voice runs for audio behavior; verify actual tool effects separately.

PLANS & PRICING

Vapi pricing

Usage-based billing includes hosting plus selected model, transcriber, voice and telephony costs. The current calculator shows USD 50 hosting for 1,000 minutes, equivalent to USD 0.05 per minute for that component only. New users are offered USD 5 starting credits. Core is an optional USD 29 monthly Success Package; usage charges continue. Legacy customers can have different terms. Do not interpret the hosting component or support-package fee as the complete cost of a call.

Pricing updated 2026-10-26. Prices may change.View current pricing →
MORE AI TOOLS

Vapi alternatives

Choose according to the audio task and the workflow you need to finish.

Integrated prompt and flow agents

Retell AI

Retell AI combines prompt or flow building with call operations. Compare tool connections, testing and the complete configured call cost.

Voice production and transcription

ElevenLabs

ElevenLabs suits generated speech, voice production or transcription. A speech provider is one component of an interactive voice application.

Speech APIs for a custom system

Cartesia

Cartesia suits speech APIs within a system you control. Account for the orchestration and business-action integration still needed around them.

COMMON QUESTIONS

Vapi FAQs

Is Vapi a text-to-speech model?

No. It orchestrates recognition, a language model, speech output and tools for an interactive agent. A TTS provider is one component of that pipeline.

Is USD 0.05 per minute the whole Vapi price?

No. The current calculator’s USD 50 hosting estimate for 1,000 minutes concerns hosting. Model, voice, recognition and phone/transport charges affect the complete call cost.

Does Core include unlimited calls?

No. Core is an optional Success Package for support and capacity. Usage billing remains active alongside it.

What is the difference between Assistants and Squads?

An Assistant handles a workflow with a prompt and tools. Squads coordinate specialised assistants and hand off context between them.

Does Vapi require application integration?

Dashboard setup can create a first agent, but business actions need the relevant connected tools. Advanced function-calling workflows require a server endpoint or API integration.

Do I always need my own webhook server for a tool?

Not for every integration. API Request can call an existing HTTPS API directly. Function tools fit Vapi’s webhook protocol, asynchronous execution or supported browser-side actions. Both require a valid connection and response contract for the business action.

Does a passed Chat simulation validate the voice experience?

No. Chat simulations skip speech generation and transcription. Voice simulations exercise the audio path and should be listened to for pronunciation, pauses and interruptions. A mocked tool result also needs separate verification against a real sandbox service.

Do draft changes immediately affect live calls?

The versioning guide separates unpublished drafts from the current published version. Publishing changes the current version. Restoring an assistant creates a new current version immediately; restoring a tool loads its configuration into a draft for later publication.

Is this your product? Claim this page to update your listing.