Vapi

Developer platform for building low-latency voice agents over the phone

AI Agents & AutomationFree planOverseas★★★★☆ 4.0

Visit Vapi official site ↗ View similar tools Compare alternatives →

What is Vapi?

Vapi lets developers build voice agents that answer and place phone calls. You configure a model, a transcription service and a voice provider, then write the logic that decides what the agent says and does; Vapi handles the telephony, the streaming and the interruption handling that make a conversation feel natural. It ships with building blocks such as knowledge bases, tool calls to your APIs and transfers to a human, plus webhooks and analytics. Server-side latency is the hard problem in voice, and the platform's main selling point is how it manages that pipeline.

Last updated: 2026-09-20. This site only provides an index; for exact features, pricing, and licensing, see the official website.

Key features

Inbound and outbound phone calling with managed telephony
Interruption handling and low-latency streaming pipeline
Choice of speech-to-text, model and voice providers
Tool calls that let an agent act on your own APIs
Warm transfer to a human and voicemail detection
Webhooks, call recording and analytics per assistant

Pros & cons

Strengths

Removes the hard parts of real-time voice plumbing
Provider-agnostic: swap models and voices independently
Fast to a working phone demo compared with building from scratch

Watch out for

Per-minute costs stack across several providers
Voice agents still fail on accents and noisy lines
Production quality needs careful prompt and latency tuning

Best for & use cases

phone support agents, appointment booking, outbound qualification and voice prototypes

If you're comparing similar products, check the alternatives below, or browse all tools in the AI Agents & Automation category.

FAQ

What do I actually have to build?

The assistant configuration and the logic behind its tools and responses. Telephony, streaming, interruption handling and provider switching are managed by the platform.

How does voice pricing work?

It is usage based per minute, and the underlying speech-to-text, model, voice and telephony providers bill separately. That layering makes a pilot call important before forecasting cost.

Can it transfer to a person?

Yes. Warm transfers and number routing are supported, which is usually how teams launch, letting the agent handle routine calls and hand over anything unusual.