Vapi

Vapi is a developer-first platform for building, deploying, and refining voice agents across phone and web experiences. It fits teams that need model flexibility, serious telephony controls, and testing tools rather than a purely no-code bot builder. The main decision is how much simultaneous traffic, data retention, and support your operation requires.

Overview

Vapi combines call orchestration, real-time monitoring, and enterprise configuration in one platform. It targets developers while providing the controls businesses need for production voice operations. The platform supports inbound and outbound calling, human handoff, and custom phone numbers, so it can cover customer service, sales, and internal call workflows.

Key features

Vapi says responses arrive in under 500ms, a useful performance target for natural conversational timing. Turn-taking, human-like voices, custom pronunciation, and keywords help teams shape how an agent sounds and behaves. Agents can draw on external data and trigger tools, including MCP, during live calls, which is valuable when a conversation must look up information or initiate an action.

Model choice is unusually broad. Teams can bring their own model or select from a managed catalog of more than 200 models and integrations. Telephony includes custom SIP, warm transfers, voicemail detection, phone-number management, and DTMF support. Those capabilities make Vapi better suited to real call-center and business processes than platforms limited to simple browser conversations.

Displayed integrations include OpenAI, Anthropic, Twilio, Zendesk, Salesforce, HubSpot, and Slack. Automated test suites, reporting and analytics dashboards with exports, and A/B experiments for prompts, voices, and flows support an iterative development process. Language coverage includes English, Spanish, Mandarin, and more than 100 additional languages. Trust Center resources cover CCPA, GDPR, HIPAA, SOC 2, PCI DSS v4.0.1, and SOC 3.

Pricing

The No Success Package costs free, billed $0 to start. It includes $5 in free credits, two organizations, one Vapi phone number, four concurrent calls, 14 days of raw-data retention, the full model catalog, and full platform access. Vapi hosting is prepaid at $0.05 per minute, while models are charged at pass-through cost. Discord peer support and email support without an SLA are included. This is useful for evaluation or a small prototype, but four concurrent calls and short retention leave little headroom for a busy service.

Core costs $29 a month, prepaid with auto-renewal. Models remain pass-through, and the plan provides two organizations, five phone numbers, 10 concurrent calls, 30 days of raw-data retention, ZDR, email support, and a two-business-day response SLA. It is the sensible paid starting point for a small production team that needs more capacity and a defined support commitment without enterprise pricing.

Pro is $999 a month, with billing described as 10% of Vapi hosting fees and a $999 monthly minimum. It raises capacity to 10 organizations, 10 phone numbers, 30 concurrent calls, and 180 days of retention. RBAC, ZDR, dedicated Slack support, a one-business-day response SLA, and a 99% uptime SLA suit teams with operational and governance requirements. The minimum makes Pro difficult to justify for modest usage.

Premier uses custom pricing and direct sales. It adds elevated organizations, concurrency, and retention; custom phone-number counts; priority support; a named support team and onboarding; a one-hour P0 response SLA; 99.9% uptime; custom RBAC, ZDR, and SSO; and dedicated SIP. It is aimed at larger deployments that need contractual service levels and tailored infrastructure.

Platforms

Vapi supports Android, iOS, web, and an API. The API emphasis makes it practical for teams embedding agents into existing products and telephony systems rather than relying on a single standalone client.

Who it's for

Choose Vapi when developers need to control models, prompts, tools, SIP, phone numbers, and human escalation in one stack. Core fits smaller production deployments; Pro fits organizations needing higher concurrency, longer retention, role controls, and faster support. The free-start package is best for experiments that can tolerate limited simultaneous calls and 14-day raw-data retention.

Pros and cons

• Pros: More than 200 model options plus bring-your-own-model support; broad telephony with warm transfer, SIP, DTMF, and voicemail detection; multilingual coverage beyond 100 languages; automated testing, analytics, exports, and A/B experiments; documented enterprise compliance resources.
• Cons: Free usage is constrained to four concurrent calls and 14 days of retention; model costs pass through on every plan; Pro carries a $999 monthly minimum; free-start support has no SLA, while Premier requires a sales-led custom arrangement.

Alternatives

LiveKit Agents is a better fit when you want freemium pricing, self-hosting, and broad desktop and server platform coverage, with a Build plan offering 1,000 agent-session minutes and five concurrent sessions.

Rapida AI suits teams preferring to bring their own STT, TTS, LLM, and SIP, with a self-hosted orientation and community support.

Voiceflow is worth considering for agency workflows that need multi-client workspace management and white-labeling.

Retell AI is an alternative for usage-based voice-agent pricing, with a pay-as-you-go plan and 20 concurrent calls included.

Ring AI fits simpler testing needs, including a free tier with 100 minutes per month.

Tone is suited to self-hosted deployments with unlimited agents, calls, and workflows and bring-your-own infrastructure.

Yellow.ai is an option for teams wanting a web, iOS, and Android service with a free plan capped at one AI agent and 500 included monthly interactions.

Intervo is the alternative to consider when open-source self-hosting and MIT-licensed source access are priorities.

Browse the AI Voice Agents and Conversational AI Platforms rankings for more options.

Verdict

Vapi is a strong choice for developer-led teams building production voice agents that need flexible models, deep telephony, multilingual conversations, and measurable iteration. Pick Core for a modest live deployment, or Pro and Premier when concurrency, retention, governance, and support commitments justify the cost. Look elsewhere if four concurrent free calls are insufficient, if pass-through model billing is undesirable, or if a $999 Pro minimum is beyond your budget.