ElevenLabs vs Vapi
Four of the nine Vapi receptionists documented on this site use ElevenLabs for the voice. That is the twist in this comparison: ElevenLabs is both a supplier inside Vapi and, through its agent platform, now called ElevenAgents, a complete alternative to it. You can build a phone agent with an ElevenLabs voice on either one. The question is who runs the call: Vapi, which stitches together a transcriber, a model and a voice from whichever vendors you choose, or ElevenLabs, which runs the whole stack itself and charges one rate per minute.
Short answer: pick ElevenAgents if you are an owner or a small agency that wants one bill, a predictable per-minute price, a visual workflow builder, built-in testing and web, WhatsApp and SMS channels alongside the phone. Pick Vapi if you are a builder who wants to choose every component, needs a free US number to start, or wants the lowest platform fee and is happy to manage the pass-through costs that come with it. On total cost at 1,000 minutes a month the two land within a few dollars of each other once the model and telephony are added, so the decision is about control and operations, not price.
Facts are as of November 21, 2026, from ElevenLabs' agents pricing page and docs and from Vapi's pricing page and docs. Operator cost figures are self-reported in this site's listings. Nothing here was tested hands-on by this site.
The Short Answer
ElevenLabs is the better choice for an owner or small agency that wants one vendor, one per-minute rate and built-in channels and testing. Vapi is the better choice for a builder who wants to pick each component and pay the lowest platform fee.
Put another way: ElevenLabs suits a shop owner, clinic or agency that wants to ship a receptionist on one bill and keep it running, while Vapi suits a developer or technical agency that wants component-level control and the cheapest platform layer.
| Feature | ElevenLabs | Vapi |
|---|---|---|
| Best for | an owner or small agency that wants one vendor, one per-minute rate and built-in channels and testing | a builder who wants to pick each component and pay the lowest platform fee |
| What it is | a full voice agent platform from the company whose voices most agents already use | an orchestration layer that connects your choice of transcriber, model, voice and telephony |
| Platform price per minute | $0.08 per call minute, flat, on every paid plan | $0.05 per minute hosting fee on every call |
| What else you pay | the language model at the provider's published rate, plus your Twilio or SIP carrier | transcriber, model and voice passed through at cost, plus telephony if you import a number |
| Rough total at 1,000 minutes a month | about $100 to $140: Pro at $99 with 1,238 minutes, plus model and carrier | about $90 to $140: Vapi's own calculator estimates $82 to $129 before telephony |
| Free tier | 15 call minutes and 4 concurrent calls, no card | $5 in free credits and one free US number for inbound calls |
| Paid plans | Starter $6, Creator $22, Pro $99, Scale $299, Business $990 a month, with included minutes | pay as you go, or a Core success package at $29 a month; Pro at 10% of hosting with a $999 minimum |
| Concurrent calls | 4 free, 6 Starter, 10 Creator, 20 Pro, 30 Scale, 40 Business; burst to 3x at $0.16 a minute | 4 on pay as you go, 10 on Core, 30 on Pro; extra lines $10 each a month |
| Phone numbers | import a Twilio number or connect a SIP trunk; no numbers sold by ElevenLabs on the pages we read | one free US number, inbound only; import from Twilio, Telnyx, DIDWW or SIP for outbound |
| Outbound and batch calling | outbound through an imported Twilio number; batch calls documented | outbound through an imported number via the API; no batch feature on the pages we read |
| Models | OpenAI, Google, Anthropic and ElevenLabs-hosted open models, billed at cost; custom LLM supported | OpenAI, Anthropic, Google and others, at cost; your own model server supported |
| Voices | ElevenLabs' own library, included in the per-minute rate | 15 voice providers including ElevenLabs, Cartesia and OpenAI; ElevenLabs works through Vapi's key or yours |
| Building the agent | a dashboard with a Workflow Builder on every plan, knowledge base, tools, MCP and a CLI | a single-prompt Assistant with tools, and Squads for multi-agent handoffs; Workflows were retired on August 18, 2026 |
| Testing | automated tests, simulations, A/B experiments and conversation analysis | Evals for single decisions and Simulations in chat or voice mode |
| Channels beyond the phone | web widget, mobile, WhatsApp, SMS, email and chat | phone, web and WebRTC; other channels are your build |
| Compliance | HIPAA BAA on Enterprise; SOC 2, ISO 27001, zero retention mode and data residency listed | HIPAA add-on at $2,000 a month; zero data retention included with Core and Pro; 14-day retention on pay as you go |
| Who should pick it | a shop owner, clinic or agency that wants to ship a receptionist on one bill and keep it running | a developer or technical agency that wants component-level control and the cheapest platform layer |
| Documented builds on this site | 6 listings use ElevenLabs, almost all as the voice inside Vapi or Retell; none yet on ElevenAgents as the platform | 9 listings |
What The Table Is Really Telling You
On what it is, ElevenLabs leans toward a full voice agent platform from the company whose voices most agents already use, while Vapi leans toward an orchestration layer that connects your choice of transcriber, model, voice and telephony.
On platform price per minute, ElevenLabs leans toward $0.08 per call minute, flat, on every paid plan, while Vapi leans toward $0.05 per minute hosting fee on every call.
On what else you pay, ElevenLabs leans toward the language model at the provider's published rate, plus your Twilio or SIP carrier, while Vapi leans toward transcriber, model and voice passed through at cost, plus telephony if you import a number.
ElevenLabs for AI Workflows
ElevenLabs started as the voice most builders reach for and has turned that into a complete agent platform. ElevenAgents gives you a dashboard where you create an agent, pick a model, write its instructions, upload a knowledge base, attach tools and webhooks, and connect it to a phone number, a web widget, WhatsApp, SMS, email or chat. A Workflow Builder for branching conversations is included on the free plan. The company's docs list automated tests, simulations, A/B experiments and conversation analytics, plus a hosted MCP server and a CLI for managing agents as code. There is no limit on the number of agents on any plan.
The pricing is the simplest in this category. Every paid plan charges $0.08 per call minute, and the plan's included minutes are priced at the same rate: Starter at $6 a month includes 75 minutes, Creator at $22 includes 275, Pro at $99 includes 1,238, Scale at $299 includes 3,738 and Business at $990 includes 12,375, as of November 2026. The free plan includes 15 minutes and 4 concurrent calls. That per-minute rate covers transcription, orchestration and the ElevenLabs voice. The language model is billed separately at the provider's published rate with no markup, and ElevenLabs says it charges nothing extra for telephony; your Twilio or SIP carrier bills its own minutes. Go over your plan's concurrency and burst pricing applies at $0.16 a minute, up to three times your limit.
The limits are worth knowing. ElevenLabs did not sell phone numbers on any page we read: you import a Twilio number, which supports inbound and outbound, or connect a SIP trunk. Concurrency is modest on the cheap plans, with 6 simultaneous calls on Starter and 10 on Creator, which is plenty for one location and tight for an agency running several. A HIPAA business associate agreement is listed on the Enterprise plan only, so a clinic that needs one is looking at a sales conversation, not a $99 plan. And the model list changes: the docs note that two Gemini models are being removed on November 30, 2026, with conversations moved to newer ones, so an agent's behavior can shift under you if you do not pin and re-test.
No build on this site documents ElevenAgents as the platform yet. The six listings that use ElevenLabs use it as the voice inside Vapi, including A $3,000 AI Voice Receptionist for Salons and Barbershops Built With Vapi and Make (builtwithagents.ai/strategy/salon-ai-receptionist-3000-vapi-make) and A Developer Built an AI Receptionist for Her Brother's Mechanic Shop to Recover 100+ Missed Calls a Week (builtwithagents.ai/strategy/ai-receptionist-mechanic-shop-missed-calls-rag), or inside Retell, as in AI Voice Receptionist for HVAC Businesses: Full Build From Scratch Using Retell AI (builtwithagents.ai/strategy/ai-voice-receptionist-hvac-retell-ai-full-build). Those builds show what the voice sounds like in production. They do not show ElevenLabs' own orchestration, transcription or tool-calling under load, which is the part you would be betting on.
Vapi for AI Workflows
Vapi is the platform behind more documented receptionists on this site than any other: nine listings, covering plumbing, pest control, salons, real estate, a mechanic shop, veterinary clinics and patient intake. Its model is to run the call and let you choose everything else. Each assistant is a transcriber, a language model and a voice, from a list of providers that includes OpenAI, Anthropic, Google, Deepgram and ElevenLabs, plus your own model server if you have one. Tools connect to your APIs, and this is why so many builds pair Vapi with n8n or Make for calendar lookups and CRM logging, as in An Emergency Plumbing AI Voice Agent Built in Three Steps With Vapi and n8n (builtwithagents.ai/strategy/emergency-plumbing-voice-agent-vapi-n8n) and A Plumbing AI Receptionist That Books Emergency Calls and Logs Everything to a CRM (builtwithagents.ai/strategy/plumbing-ai-receptionist-vapi-make-calcom).
The price is a $0.05 per minute hosting fee plus pass-through costs. Vapi's pricing page shows the components at cost, with example ranges of roughly $0.01 a minute for Deepgram transcription, $0.008 to $0.045 for an OpenAI model and $0.015 to $0.024 for an ElevenLabs voice, and its calculator estimates $82 to $129 a month for 1,000 minutes before telephony. Telephony is free over Vapi's own SIP and WebRTC transport; an imported Twilio number adds Twilio's $0.008 inbound and $0.014 outbound rates. New accounts get $5 in free credits and one free US phone number, which is inbound only. Pay as you go includes 4 concurrent calls and 14 days of data retention. The Core success package at $29 a month raises that to 10 concurrent calls, 5 numbers, 30 days of retention, zero data retention included and a two-business-day support response. Pro costs 10 percent of your hosting spend with a $999 a month minimum. Extra concurrent lines are $10 each a month, and the HIPAA add-on is $2,000 a month.
What Vapi asks of you is attention. Three vendors' prices move independently, and the per-minute total depends on choices you make in the dashboard, so budgeting takes a spreadsheet rather than a plan page. Outbound calling needs an imported number. And the builder surface changed this year: Vapi retired its visual Workflows on August 18, 2026, and points builders to Squads, which hand a call between assistants with context, so a build that followed an older tutorial may need rework. On the other side, Vapi's testing is real, with Evals that check a single decision and Simulations that hold a whole conversation in chat or voice, and its docs are written for people who will read them.
The operators on this site are candid about Vapi's failure modes. A Voice AI Receptionist for Veterinary Clinics That Books Appointments Automatically (builtwithagents.ai/strategy/voice-ai-receptionist-veterinary-clinics) reports about 60 percent accuracy on appointment handling in early testing, with name and number capture and invented slots as the causes. Nate Herk Built an AI Voice Agent That Calls Every New Lead Within Seconds of Their Form Submission (builtwithagents.ai/strategy/nate-herk-ai-voice-agent-calls-every-new-lead-n8n-vapi) shows the outbound pattern with an ElevenLabs voice and n8n. None of those problems are unique to Vapi. They are what happens when the builder, not the platform, owns the integration, and that is the deal Vapi offers.
Our Verdict
Pick ElevenAgents if you want to ship a receptionist and move on. One per-minute rate, included minutes on a plan you can read in a minute, a workflow builder and tests on the free tier, and web, WhatsApp and SMS channels that a Vapi build would need you to construct. Start on Creator at $22 a month, which covers 275 minutes and 10 concurrent calls, bring a Twilio number, and use the 15 free minutes to run the test script from this site's guide to testing a voice agent before you pay anything. If you are a clinic that needs a HIPAA agreement, ask sales before building, because it is Enterprise-only.
Pick Vapi if you or your agency will own the integration anyway. It is cheaper at the platform layer, the free inbound number means a working demo costs nothing, every component is swappable, and the nine documented builds on this site are a library of working patterns with n8n and Make. Buy the $29 Core package once you have a client, for the 10 concurrent lines and the zero-retention setting, and keep the testing habit, because the vet clinic listing shows what a Vapi build looks like before it has one. If the choice is between Vapi and Retell rather than ElevenLabs, see builtwithagents.ai/blog/retell-vs-vapi-small-business.
If you are already on Vapi with an ElevenLabs voice and it works, there is no reason to move. The voice is the same on both. Switch only when you want the channels, the one-bill pricing or the included testing, and run both in parallel on a test number for a week before you forward the real one.
FAQ
Is ElevenLabs or Vapi cheaper for a voice agent?
At the platform layer Vapi is cheaper at $0.05 a minute against ElevenLabs' $0.08, but Vapi's rate excludes the transcriber and voice, which ElevenLabs includes. Vapi's own calculator estimates $82 to $129 a month for 1,000 minutes before telephony; ElevenAgents' Pro plan is $99 for 1,238 minutes plus the model at cost and your carrier. In practice they land within a few dollars of each other as of November 2026.
Can I use ElevenLabs voices in Vapi?
Yes. ElevenLabs is one of Vapi's 15 voice providers, and Vapi's docs say you can use it through Vapi's default integration or connect your own ElevenLabs account. Four of the nine Vapi builds on this site use an ElevenLabs voice.
Does ElevenLabs provide phone numbers?
Not on the pages we read. ElevenAgents connects to a number you import from Twilio, which supports inbound and outbound calls, or to a SIP trunk. ElevenLabs says it charges nothing extra for telephony; your carrier bills its own minutes.
Does Vapi give you a free phone number?
Yes, one free US number per account, and you can request it without adding a payment method. It is inbound only. For outbound calls you import a number from Twilio, Telnyx or DIDWW or connect a SIP trunk.
Which is better for HIPAA?
Neither is cheap. ElevenLabs lists a business associate agreement on its Enterprise plan only. Vapi sells a HIPAA add-on at $2,000 a month on any plan. A single clinic should price both against a managed healthcare receptionist product before building.
What happened to Vapi Workflows?
Vapi retired Workflows on August 18, 2026; existing workflows stopped running the next day and the dashboard no longer creates new ones. Squads, which pass a call between assistants while keeping context, are the replacement. ElevenAgents still includes a visual Workflow Builder on every plan.
Which one has better testing tools?
Both have built-in simulation now. ElevenAgents lists automated tests, simulations and A/B experiments. Vapi has Evals, which check one decision, and Simulations, which run a full conversation in chat or voice mode. Either is enough to run the 30-call script in this site's voice agent testing guide.
External Links
Related Strategies
Real workflows on this site that use one or both of these tools.
A voice AI agent for healthcare clinics that qualifies patients, verifies insurance, and books appointments without human staff involvement.
An AI voice receptionist for a plumbing company that books emergency service calls, checks real time calendar availability, and logs every conversation to Airtable automatically.
Brendan walks through building a complete AI voice receptionist for an HVAC company — from call flow diagram to production-ready agent that books appointments, handles emergencies, and transfers calls automatically