Vapi
The developer platform for production voice AI agents — phone calls that resolve issues, not phone trees.
About
Vapi is the infrastructure layer for building and deploying AI voice agents that handle real phone calls: customer support, appointment scheduling, outbound sales, and lead qualification. You configure an assistant (model, voice, tools, and call logic) in the dashboard or API, attach phone numbers, and Vapi handles telephony, low-latency audio streaming, and integrations — it is model-agnostic, so you can swap the underlying LLM without rebuilding the phone stack.
The traction is the story here: a $50M Series B at a ~$500M valuation in May 2026 led by Peak XV with Microsoft's M12 participating, 1M+ developers and 2.7M+ agents built on the platform (provider-reported), and Amazon Ring routing 100% of its inbound support calls through Vapi after evaluating 40+ vendors. It is closed SaaS with usage-based pricing, so the open-source crowd should look at Pipecat instead — but for teams that want voice agents without operating telephony infrastructure, Vapi is the current enterprise reference deployment.
What it can do
- Configurable voice agents: model, voice, prompt, and tools per assistant
- Inbound and outbound telephony with low-latency audio streaming
- Model-agnostic: swap LLMs without rebuilding the phone stack
- Function calling, RAG, and CRM integrations during live calls
- Call analytics, transcripts, and non-engineer-friendly tuning
- Enterprise scale: 1M-5M calls/day on the platform (provider-reported)
Setup
- Sign up at vapi.ai and create an assistant in the dashboard.
- Configure the model, voice, system prompt, and tools (function calling, RAG, integrations).
- Attach a phone number (import your own or provision through Vapi).
- Test with web calls, then go live on inbound/outbound telephony.
- Monitor calls, transcripts, and analytics; tune prompts without engineering help.
Prerequisites
A Vapi account; a phone number (yours or provisioned); provider keys only if you bring your own models.
Use cases
Works with
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: official site and docs verify the assistant/phone-number model and model-agnostic architecture; independent May 2026 coverage corroborates the $50M Series B at ~$500M (Peak XV-led, M12 participating) and the Amazon Ring deployment (100% of inbound calls after a 40+ vendor evaluation). Strengths: the current enterprise reference for voice agents, real Fortune-500 validation. Docked for: (1) closed SaaS, no public repo — adoption is a proxy estimate; (2) call-volume figures are provider-reported; (3) usage-based pricing adds up at scale. Adoption: thinner evidence (SaaS proxy).
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 8.0/10
- Openness · license + docs
- 5.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · Adoption 7: SaaS proxy estimate — $50M Series B at ~$500M valuation (May 2026, Peak XV-led, M12 participating), Amazon Ring 100% inbound deployment, 1M+ developers per independent coverage; thinner evidence (no public repo metric)
- · Maintenance 8: active docs, changelog, and platform (provider-observed); no repo recency signal available
- · Openness 5: closed SaaS (0.7*3) + official docs verified at docs.vapi.ai (0.3*10)
- · Manual 7: official site/docs verify the assistant and phone-number model; independent coverage corroborates funding and Ring deployment; docked for closed SaaS, provider-reported call volumes, usage-based pricing
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.