ChipChat News
Loading…

Vapi

Vapi lets developers build voice AI agents for support, sales and scheduling, combining their choice of speech, LLM and voice models with telephony, testing and monitoring.

Vapi website screenshot
Image: vapi.ai

What is Vapi?

Vapi is a developer platform for building, testing and running voice AI agents that hold real phone or web conversations. It handles the plumbing between speech recognition, a language model, a synthetic voice and telephony, so teams can focus on the agent's behaviour instead of real-time audio infrastructure. Typical deployments include customer support lines, lead qualification and appointment booking.

Vapi is aimed at engineers and product teams, from solo builders to large enterprises, and is designed to be configured through its API and dashboard rather than as a no-code chatbot for non-technical users.

Key Features of Vapi

Bring-your-own model stack

You choose the transcriber, language model and voice provider for each agent, for example Deepgram, OpenAI and ElevenLabs. Model usage is passed through at provider cost without a markup.

Telephony and transports

Agents can answer and place calls over Vapi's own numbers and SIP, WebSockets or WebRTC, or connect to carriers such as Twilio, Vonage and Telnyx, whose fees are charged by the provider.

Testing and simulation

Built-in simulation lets you run test calls against an agent before it reaches customers, which helps catch broken flows and prompt issues early.

Monitoring and observability

Call monitoring surfaces problems in live traffic, and data retention windows grow with each package, so teams can review transcripts and improve agents over time.

Enterprise controls

Vapi lists SOC 2, HIPAA and PCI compliance, SSO, OAuth2, role-based access control, conversation guardrails and zero data retention options, with HIPAA handling available as a paid add-on.

Vapi Pricing

Vapi pricing has two parts: usage, billed per minute (Vapi's own hosting fee plus pass-through model and transport costs), and an optional success package that adds support, limits and compliance features. Its calculator estimated roughly $82 to $129 per month for 1,000 minutes with Deepgram, OpenAI and ElevenLabs, depending on the model chosen. Add-ons include extra concurrency at $10 per line per month, extra organizations at $20 each per month and HIPAA compliance at $2,000 per month.

Prices checked on October 28, 2026; check the official pricing page for current plans.

Who Should Use Vapi?

  • Developers building inbound support or outbound calling agents on top of their own systems.
  • Startups that want to prototype a voice product quickly and pay only for minutes used.
  • Healthcare, finance and other regulated teams that need compliance options and data controls.
  • Enterprises running high call volumes that need SLAs and dedicated support.

Is Vapi Worth It?

For technical teams, Vapi removes much of the hardest work in voice AI, namely low-latency orchestration, telephony and monitoring, while leaving full control over models and prompts. The free starting tier and per-minute billing make it easy to experiment before committing.

It is less suited to people looking for a point-and-click receptionist, because getting good results still requires configuring models, prompts and integrations. Costs can also be hard to predict, since the final bill combines Vapi's hosting fee with transcription, model, voice and carrier charges, and the jump from Core to the $999 Pro minimum is steep. Running the pricing calculator against your expected volume is the best way to judge the fit.