Vapi vs Retell · October 2026
Vapi vs Retell AI, compared from their own pages
Vapi and Retell AI are the two self-serve voice platforms most teams shortlist. Both price per minute in US dollars, both connect to your own SIP trunk, both have a REST API and a hosted MCP server. They differ in how the bill is built, how much comes ready-made, where your data can live and what developers get. Below: the comparison line by line, the same receptionist workload priced on both, when to pick which, and VoiceDock as a third option for European teams.
Checked on vapi.ai, docs.vapi.ai, retellai.com and docs.retellai.com on 21 October 2026
Pay-as-you-go on both. Sources at the bottom of the page.
The short answer
Which one fits
Pick Vapi if
you want the lowest platform fee with models passed through at cost, you are happy to choose and tune providers yourself, and you want SDKs for web and mobile plus a CLI. US callers can be tested on a free inbound number.
Pick Retell AI if
you want more of the toolbox ready-made: a visual conversation flow builder, knowledge bases, simulation testing, 20 concurrent calls included, US and Canadian numbers at $2, and a BAA without an add-on fee.
Look at VoiceDock if
you are a European team that needs the platform layer in the EU without an enterprise package, wants a DPA at no extra cost and a contract in euros, or wants the engineers to build the line with you and sell it on under your own brand. We compare ourselves with Vapi and with Retell AI in full on separate pages.
Side by side
Vapi, Retell AI and VoiceDock, line by line
Every Vapi and Retell cell comes from that vendor's own pricing page, documentation or homepage. Every VoiceDock cell is something you can check on this site, in our docs or in the Trust Center.
| Vapi | Retell AI | VoiceDock | |
|---|---|---|---|
| Platform fee | $0.05 per minute hosting | $0.055 per minute voice infrastructure | € 0,07 per minute orchestration, for inbound, outbound and web calls alike |
| Models, voice, transcription | Passed through at cost; your own provider keys supported | Retell's own per-minute price per model and voice; a custom LLM over your own server | At cost on our keys, or your own keys at the same fee |
| The vendor's own all-in example | $82 to $129 per 1,000 minutes for its default stack, before carrier charges | $0.11 per minute before telephony; headline range $0.07 to $0.31 | We publish the components rather than one all-in figure: see pricing |
| Packages and plans | Pay-as-you-go; Core $29 a month, Pro with a $999 monthly minimum, Premier on request | Pay-as-you-go; Enterprise on request | None; prepaid credit, invoicing by arrangement |
| Concurrency | 4 on pay-as-you-go, 10 on Core, 30 on Pro; $10 per extra line per month | 20 included; $8 per extra call per month; $0.10 per minute surcharge on burst calls | Concurrent calls under fair use, more by arrangement |
| Built-in numbers | Free Vapi numbers: US area codes, inbound only | US and Canada, $2 a month, $0.015 per minute | Your own number; managed numbers on request |
| Own SIP trunk and carrier | SIP trunking; import guides for Twilio, Telnyx and DIDWW; inbound SIP URI | Elastic SIP trunking (recommended) or SIP URI dialing; guides for Twilio, Telnyx and Vonage; no Retell telephony charge | Your own number over your own SIP trunk, set up with you |
| Transfers | Blind transfer by default; warm transfer modes on Twilio calls | Cold, warm and agentic warm transfer, on phone calls only | Warm and cold transfer to a phone number or through your own SIP trunk |
| Published latency | "<500ms" average latency, on its homepage | "as low as 600ms" from the end of speech, in its docs | No single figure; where the milliseconds go on a phone call is written up instead |
| EU hosting | EU region on the Premier package only | Retell says it does not currently operate services within the EU | Orchestration, dashboard, API and storage on EU infrastructure |
| Data processing agreement | Referred to in Vapi's trust center; a standard DPA and BAA are part of the $2,000 a month HIPAA add-on | Self-sign DPA with EU standard contractual clauses, no fee | Standard DPA under GDPR Article 28, at no additional cost |
| Certifications claimed | SOC 2 Type II; HIPAA as a paid add-on; PCI DSS listed in its trust center | SOC 2 Type 1 and Type 2; HIPAA with a BAA at no fee; ISO 27001 on its homepage | Held by our infrastructure providers, listed per provider in the Trust Center |
| White-label and reselling | Solution partner programme; extra organisations at $20 a month; a third-party white-label layer is listed in its docs | Partner programme; one workspace per client recommended, with no API to create workspaces | White-label and reseller setup: a parent organisation covers its customer organisations, at your own price per customer |
| API and SDKs | REST API; client SDKs for web, iOS, Flutter, React Native and Python; server SDKs in several languages; a CLI | REST API; TypeScript and Python SDKs and a web client SDK | REST API, a TypeScript SDK and signed webhooks |
| MCP | Hosted MCP server for assistants, calls and phone numbers, with an API key; MCP tools during calls | Hosted MCP server for agents, calls, numbers, knowledge bases and more, with an API key; MCP during calls | Hosted MCP server to build and manage assistants, with OAuth sign-in; your own MCP servers as tools during a call |
| How agents are built | Assistants and Squads; Vapi retired its Workflows in August 2026 | Single and multi-prompt agents, plus a drag-and-drop Conversation Flow canvas | Assistants with tools, set up in the dashboard, over the API or over MCP |
| Contract and currency | US dollars | US dollars | Flireo B.V. in Deventer, in euros |
Same workload, every bill
1,000 calls of three minutes, priced on each
An inbound AI receptionist at 3,000 minutes a month, on each vendor's own default stack. What each total includes is in the right-hand column.
| Per month | What it includes | |
|---|---|---|
| Vapi, pay-as-you-go | $245 to $387 | Hosting $150, plus transcriber, model and voice at cost (Vapi's calculator defaults); free transport on a US Vapi number. Outside the US, carrier charges come on top: $24 for the inbound minutes with Twilio |
| Retell AI, pay-as-you-go | $377 | Voice infrastructure $165, model $120 and voice $45 (Retell's calculator default), telephony $45 and $2 for a Retell US number |
| VoiceDock | € 210 plus models and telephony | Orchestration at € 0,07; model usage at cost on our keys or billed on yours; telephony per leg at the carrier's rate |
| ElevenLabs Agents, Scale plan | $299 plus LLM usage | 3,738 call minutes included; LLM usage comes on top; your carrier bills the call |
On the platform fee alone, Vapi ($150) and Retell ($165) are both below our € 210 at this volume. Vapi and VoiceDock both pass models through at cost, so between those two the gap is the fee and what comes with it; Retell's total moves with its own price list. Which AI receptionist setup fits is a separate question from the price, and how to work out a voice agent's cost per minute applies to all three.
When to pick which
Where each one is the better choice
A comparison that crowns one winner is not worth reading. And where VoiceDock is not the answer: a team that wants to start alone tonight without ever talking to a vendor is better served by Vapi or Retell.
Vapi is the better choice when
You want the lowest fee and full control of the stack
$0.05 a minute with models at cost and your own provider keys supported. No platform fee in this comparison is lower, as long as you choose, tune and debug the providers yourself.
You build for web and mobile
Client SDKs for web, iOS, Flutter and React Native, server SDKs in several languages, and a CLI that Vapi says does everything the dashboard does.
You test with US callers
A free US number for inbound calls and $5 of credit give you a working line without a carrier account.
Retell AI is the better choice when
You want a visual flow builder
Retell's Conversation Flow is a drag-and-drop canvas, and Vapi retired its Workflows in August 2026. If your team designs calls as diagrams, Retell has the tool.
You need concurrency and a toolbox included
Twenty concurrent calls per workspace at no charge, plus knowledge bases, simulation testing and quality assurance that you switch on when you need them.
You need a BAA without an add-on
Retell signs BAAs and DPAs at no extra fee and holds SOC 2 Type 1 and Type 2. Vapi's HIPAA add-on is $2,000 a month.
VoiceDock is the third option when
The platform layer has to be in the EU
Vapi's EU region needs its Premier package, and Retell does not operate services in the EU. Our orchestration, dashboard, API and storage run on EU infrastructure, and why that matters for voice has its own page.
You want it built with you
We are the engineers who maintain the orchestration layer. We build your line in working sessions and read the SIP logs when a call fails; how that works for agencies has its own page.
You sell voice AI under your own brand
A parent organisation covers its customer organisations at your own price per customer, and telecom and VoIP providers run it on lines they already provide. What happens when a server fails is on the reliability page.
And ElevenLabs?
Vapi vs ElevenLabs Agents
ElevenLabs comes up as often as Retell, from the voice side. Its agent platform prices differently from both: a monthly plan with included call minutes.
| Vapi | ElevenLabs Agents | |
|---|---|---|
| Pricing model | $0.05 per minute hosting plus providers at cost | Monthly plans with included minutes, from $6 for 75 minutes to $990 for 12,375; extra minutes $0.08; LLM usage on top |
| Telephony | Free transport on Vapi numbers; other carriers bill you directly | No ElevenLabs telephony fee; native Twilio integration and SIP trunking, and your carrier bills you |
| Concurrency | 4 to 30 by package; $10 per extra line per month | 4 to 40 by plan; burst calls at twice the minute rate |
| EU data residency | EU region on the Premier package | An Enterprise feature |
| MCP | Hosted MCP server, with an API key | Hosted MCP server, with OAuth; MCP tools in agents, off by default |
| SDKs | Web, iOS, Flutter, React Native and Python clients | Python, JavaScript, React, Swift and Kotlin |
ElevenLabs fits when a plan with included minutes suits your volume and the voice is what you are buying: its Scale plan at $299 includes 3,738 minutes, which covers the receptionist workload above before LLM usage. Vapi fits when you want to choose every provider and pay for minutes as they come.
Questions
What people ask when they compare Vapi and Retell
Is Vapi or Retell cheaper?
On the platform fee Vapi is lower: $0.05 a minute against Retell's $0.055. The total depends on the stack: Vapi passes models through at cost, Retell charges its own price per model and voice. For 1,000 three-minute calls, Vapi's calculator defaults come to $245 to $387 a month and Retell's to $377 including a Retell number, by our arithmetic on their published rates.
Which has lower latency, Vapi or Retell?
Each publishes its own figure: Vapi shows "<500ms" average latency on its homepage, Retell's docs say "as low as 600ms". They are not measured the same way, and on a phone line much of the delay sits outside the model, as set out in the latency budget of a phone call.
Do Vapi and Retell host data in the EU?
Vapi offers an EU region, available only on its Premier package. Retell's compliance page says it does not currently operate services within the European Union. VoiceDock runs orchestration, dashboard, API and storage on EU infrastructure.
Can I bring my own SIP trunk to Vapi or Retell?
Yes, to both. Vapi documents SIP trunking and number imports from Twilio, Telnyx and DIDWW; Retell recommends elastic SIP trunking and charges nothing for telephony over your own trunk. On VoiceDock your own number runs over your own SIP trunk as well.
Do Vapi and Retell have an MCP server?
Both run a hosted MCP server that you connect with an API key: Vapi's covers assistants, calls and phone numbers, Retell's covers agents, calls, numbers, knowledge bases and more. Ours uses OAuth sign-in, so there is no key to paste.
Can I white-label Vapi or Retell?
Neither publishes a white-label programme of its own on the pages we read. Vapi has a solution partner programme and lists a third-party white-label layer in its docs; Retell has a partner programme and recommends one workspace per client. Our white-label setup gives you a parent organisation over your customers' organisations, at your own price per customer.
Vapi vs ElevenLabs: what is the difference?
Vapi charges per minute with providers at cost; ElevenLabs Agents sells monthly plans with included call minutes, from $6 for 75 minutes to $990 for 12,375, with LLM usage on top. Both connect to your own carrier over SIP or Twilio.
Sources, checked 21 October 2026
- Vapi homepage
- Vapi pricing
- Vapi: pricing and success packages
- Vapi: call concurrency
- Vapi: free telephony
- Vapi: SIP
- Vapi: warm transfer
- Vapi: EU region
- Vapi: SOC 2
- Vapi: HIPAA
- Vapi trust center
- Vapi partnerships
- Vapi: Vapify (third-party white-label)
- Vapi: MCP server
- Vapi: CLI
- Vapi: SDKs and resources
- Vapi: provider keys
- Vapi: Workflows migration
- Vapi on GitHub
- Retell AI homepage
- Retell AI pricing
- Retell AI: estimated latency
- Retell AI: concurrency and burst
- Retell AI: phone numbers
- Retell AI: custom telephony
- Retell AI: transfer call
- Retell AI: security and compliance
- Retell AI partner programme
- Retell AI: workspaces
- Retell AI: API overview and SDKs
- Retell AI: MCP server
- Retell AI: Conversation Flow
- Retell AI: custom LLM
- ElevenLabs Agents pricing
- ElevenLabs: data residency
- ElevenLabs: SIP trunking
- ElevenLabs: hosted MCP server
- ElevenLabs: burst pricing
Vendor pages only; no third-party reviews. Latency figures are each vendor's own claim, not a measurement of ours.
Comparing for a European team?
Thirty minutes, no deck. Bring your shortlist, your call volume and where your callers are, and we tell you plainly whether Vapi, Retell or VoiceDock fits.