Vapi vs Retell AI: which voice agent platform fits your build
Vapi and Retell AI are both self-serve platforms for AI phone agents, billed by the minute with no annual contract. The split is how much of the stack you assemble. Vapi charges $0.05 per minute for hosting and passes transcriber, model, voice and telephony costs through at provider rates, with $5 in free credits. Retell quotes $0.07 to $0.31 per minute all-in, with $10 in free credits and 20 concurrent calls included. Both pricing pages were checked on 20 November 2026.
Vapi suits developers who want to pick every provider, bring their own API keys and keep agent configuration in code. Retell suits teams that want more built in: conversation flow builders, warm transfer, batch calling and SIP routes into major contact-center platforms.
This page compares pricing, concurrency, telephony, testing, compliance and setup using each vendor's pricing page and docs. For more options, see our AI voice agent software category.
Vapi
Retell AI
Verdict
Pick Vapi if your team writes code and wants control over each component. You choose the transcriber, model and voice from dozens of providers, and once you validate your own provider key Vapi stops billing you for that provider. Vapi's calculator puts 1,000 minutes a month at $82 to $129 on the usage-only tier. The catch is the limits: 4 concurrent calls, one inbound-only US number and 14 days of retention until you buy a Success Package, plus $2,000 a month for HIPAA.
Pick Retell AI if you want more product out of the box and more headroom before paying for capacity. Pay as you go includes 20 concurrent calls, simulation testing and API access, and calls over the limit can connect on optional burst pricing. A build of GPT 5.1, a platform voice, infrastructure and a Retell number comes to $0.125 per minute. The trade-off is that premium voices, models and add-ons push the bill up, and SSO and role-based access sit in custom-priced Enterprise.
If you want fine control over providers and will run everything through the API, start with Vapi. If you need 10 or more simultaneous calls on day one, or want transfers and campaigns without writing them, start with Retell.
- Vapi
- Choose Vapi when you want a $0.05 per minute hosting fee with providers billed at cost, your own provider keys, squads of linked assistants and configuration kept in version control.
- Retell AI
- Choose Retell AI when you want an all-in rate from $0.07 per minute, 20 concurrent calls included, built-in flow builders and transfers, and SIP routes to Five9, Talkdesk or Amazon Connect.
Choose Vapi when…
Developers and product teams building voice agents into their own software, who want to choose every provider, bring their own keys and manage assistants through the API, CLI and Git.
- Provider costs pass through with no Vapi markup: Deepgram at $0.0095 to $0.0099 per minute, OpenAI models at $0.0077 to $0.0452 and ElevenLabs voices at $0.0146 to $0.0238
- Bring-your-own provider keys: once a key is validated in the Integrations tab, that provider bills you directly and Vapi charges only hosting
- Squads link several assistants with context-preserving transfers, so a front-desk agent can hand a caller to a billing agent
- Simulations, scripted evals and call monitoring, with an open-source GitOps repository for assistants and tests
- The usage-only tier allows 4 concurrent calls, 1 phone number and 14 days of raw data retention
- The free Vapi number is a US number that takes inbound calls only, so outbound testing needs Twilio, another carrier or SIP
- When every concurrency slot is busy, new inbound and outbound calls are blocked until one frees up
- HIPAA is a $2,000 per month add-on, and SSO and the EU region need the custom Premier package
Choose Retell AI when…
Operations and engineering teams that want a self-serve agent platform with transfers, campaigns and contact-center routing built in, including clinics, lenders, agencies and call centers adding AI over SIP.
- 20 concurrent calls included on pay as you go, with extra lines at $8 each per month
- Optional burst pricing lets calls above the limit connect at an extra $0.10 per minute instead of being rejected
- Single prompt and conversation flow builders, a knowledge base, batch calling and post-call analysis in one account
- Integrations page lists Five9, NICE CXone, Talkdesk, LiveVox, Amazon Connect and RingCentral, plus HubSpot, Salesforce, Zoho CRM, Calendly and Cal.com
- The top of the $0.07 to $0.31 range is more than four times the bottom, and ElevenLabs voices alone run $0.04 to $0.10 per minute
- Add-ons stack: PII removal $0.01, knowledge base $0.005 and AI quality assurance $0.10 per minute after 100 free minutes
- Calls end after one hour by default, adjustable to two hours in agent settings
- Custom SSO, role-based access control and custom MSA and DPA terms are listed under Enterprise
At a glance
Vapi is a developer platform for voice agents on phone and web calls. An assistant is a system prompt plus tools and structured outputs; squads chain several assistants together. You configure the transcriber, model and voice from a long provider list, point server URLs at your backend to receive call events, and manage the setup through the dashboard, API, CLI or a GitOps repository. Billing is a pre-paid hosting fee plus provider costs, with optional Success Packages that raise limits and add support.
Retell AI packages more of the agent into the product. You build from a single prompt, a node-based conversation flow or a template, attach a knowledge base, add functions for booking or CRM updates, and test with simulations before attaching a number. It sells US and Canadian numbers, accepts Twilio, Vonage, Telnyx or SIP, and offers chat agents from $0.002 per message. Enterprise adds SSO, role-based access and custom contract terms.
Vapi
- $0.05 per minute hosting, pre-paid, plus providers at cost
- $5 in free credits on the usage-only tier
- Success Packages: Core $29 a month, Pro 10% of hosting with a $999 minimum, Premier custom
- 4, 10 or 30 concurrent calls by package; extra lines $10 each a month
- Outbound campaigns from a CSV of up to 10,000 contacts
Retell AI
- $0.07 to $0.31 per minute all-in for voice, from $0.002 per message for chat
- $10 in free credits and 20 concurrent calls on sign-up
- $0.055 per minute infrastructure, $0.015 per minute platform voices, $0.015 per minute telephony on Retell numbers
- Extra concurrency $8 per line per month; optional burst at $0.10 per minute
- Batch calling at $0.005 per dial, branded calls at $0.10 per outbound call
Who each is built for
Vapi is built for engineers who treat the voice agent as part of their own product. If the agent should call your backend APIs, you want to swap models or voices freely, or you plan to use your own OpenAI or ElevenLabs contracts, Vapi's pass-through billing and provider keys fit. It's a weaker match for a team without developers, since tools, webhooks and authentication are yours to build.
Retell AI is built for teams that want the call-handling logic ready-made. Clinics and lenders running scheduling or collections calls, agencies building agents for clients and contact centers sending queue traffic to AI over SIP get flow builders, warm transfer, IVR navigation and batch campaigns in one account, and 20 included lines suit a pilot with real traffic. Teams that want their own provider contracts get less of that control than on Vapi.
Vapi
Best for developer teams that want to choose and pay for each provider directly, run agents through the API and Git, and can live with tight concurrency until they buy a package.
Retell AI
Best for operations and engineering teams that want a metered all-in rate, 20 included lines, built-in transfers and campaigns, and SIP routes into existing contact-center software.
Comparison matrix
These rows cover what usually decides a self-serve voice agent purchase: how the rate is built, how many calls run at once, how calls reach your phones and which compliance terms cost extra. Figures are from each vendor's pricing page and docs, checked on 20 November 2026.
| Criterion | Vapi | Retell AI |
|---|---|---|
| Pricing model | Usage-based: $0.05/min hosting plus provider costs at cost, optional monthly Success Packages | Usage-based: per minute (voice) and per message (chat) from component rates, no platform fee |
| Starting price | $0.05/min hosting; Vapi's example totals about $0.08 to $0.13/min | $0.07/min; Enterprise custom |
| Free plan or trial | $5 in free credits, no minimum minutes | $10 in free credits |
| Deployment | Web | Web |
| API | Yes | Yes |
| Included concurrency | 4 lines usage-only, 10 on Core, 30 on Pro; $10 per extra line a month | 20 lines; $8 per extra line a month; optional burst at +$0.10/min |
| Over the concurrency limit | New calls blocked until a slot frees | Calls connect on burst pricing if enabled |
| Telephony | One free US inbound-only number; Twilio $0.008 in, $0.014 out; Vonage, Telnyx or free SIP | Retell numbers (US, Canada) at $0.015/min; Twilio, Vonage, Telnyx or SIP at no Retell charge |
| Bring your own provider keys | Yes; Vapi stops billing for that provider | Custom LLM over WebSocket; voice and infrastructure billed by Retell |
| Testing before launch | Simulations, scripted evals, monitoring, GitOps repository | Playground, simulation testing, AI quality assurance add-on |
| Compliance and access | HIPAA add-on $2,000/mo; SSO and EU region on Premier | HIPAA-ready per pricing page; SSO and RBAC on Enterprise |
Pricing and packaging
Vapi's pricing page, checked on 20 November 2026, starts with a pre-paid $0.05 per minute hosting fee and $5 in free credits. Provider costs sit on top at cost: Deepgram transcription at $0.0095 to $0.0099 per minute, OpenAI models at $0.0077 to $0.0452 and ElevenLabs voices at $0.0146 to $0.0238. Vapi telephony and SIP are free to use for transport, while Twilio is $0.008 per minute inbound and $0.014 outbound, Vonage $0.00814 and Telnyx $0.0055. Vapi's calculator puts 1,000 minutes at $82 to $129 a month: $50 hosting, $10 transcription, $8 to $45 for the model and $15 to $24 for the voice.
Vapi's Success Packages add limits and support on top of usage. Core is $29 a month with 10 concurrent calls, 5 numbers and 30 days of retention. Pro is 10% of your hosting fee with a $999 monthly minimum, with 30 lines, 180 days of retention and a 99% uptime SLA. Premier is custom and adds SSO, and the docs say the EU region is Premier-only. Add-ons are HIPAA at $2,000 a month, extra lines at $10 and extra organizations at $20, each a month.
Retell AI's pricing page, checked the same day, quotes $0.07 to $0.31 per minute for voice and from $0.002 per message for chat, with $10 in free credits and 20 concurrent calls. The rate sums $0.055 per minute infrastructure, a voice ($0.015 for platform voices such as Cartesia and OpenAI, $0.04 to $0.10 for ElevenLabs), an LLM (GPT 5 nano $0.0016, GPT 5.1 $0.04) and $0.015 per minute on a Retell number, with no charge for SIP. GPT 5.1 with a platform voice on a Retell number totals $0.125 per minute, or $125 for 1,000 minutes. Extra lines are $8 a month; add-ons include knowledge base retrieval at $0.005 per minute, PII removal at $0.01 and AI QA at $0.10 after 100 free minutes.
At 1,000 minutes a month the two land close together, so concurrency often decides the cost. Retell includes 20 lines. On Vapi, 20 lines means Core at $29 plus 10 extra lines at $10, or $129 a month before usage. Price both at your peak hour with the model and voice you'll run, plus Retell's add-ons or Vapi's HIPAA fee if they apply.
Setup and operations
Vapi's quickstarts cover inbound, outbound and web calls: a system prompt, a model preset and a phone number. After that the work is engineering: tools that call your APIs, server URLs that receive call events, secured endpoints and provider keys. The one free number is inbound only and can't be ported out, so import a carrier number or set up SIP before testing outbound campaigns.
Retell AI setup follows build, test, deploy and monitor. You create an agent from a prompt, flow or template, add a knowledge base and functions such as Cal.com booking, then test with simulations before buying a number or connecting Twilio, Vonage, Telnyx or a SIP trunk. The API and Node.js and Python SDKs cover the same steps.
Concurrency behaves differently at the limit. Vapi shares slots across every organization on a subscription, and when they're full, calls are blocked. Retell sets 20 lines per workspace; with burst enabled, extra calls connect at $0.10 more per minute, up to the lower of three times your limit or your limit plus 300. Retell calls also end after one hour by default, adjustable to two.
Plan compliance before the first live call. On Vapi, HIPAA is switched on organization-wide from the dashboard, needs a signed BAA and HIPAA-compliant providers, and turns off zero data retention; EU organizations run on separate endpoints such as api.eu.vapi.ai. Retell's pricing page calls the platform HIPAA-ready but doesn't clearly show which plan carries BAA terms, so confirm that with sales before any patient calls.
Before you choose
Answer these with whoever will build the agent and whoever owns the phone bill before loading credits.
- How many calls will run at the same time in your busiest hour, and should overflow calls be blocked or connected at a higher rate?
- Do you already pay OpenAI, ElevenLabs or Deepgram directly, and would you rather keep those contracts than buy through the voice platform?
- Will a developer own tools, webhooks and authentication, or do you need flow builders and transfers that work without much code?
- Which phone system carries your calls today: Twilio, Telnyx, Vonage, a SIP trunk or a contact center such as Five9 or Talkdesk?
- Do calls involve protected health information, and does procurement need a BAA, SSO or EU data hosting from the start?
- How long is your longest real call, and how many outbound numbers will campaigns need?
Frequently asked questions
Short answers to the questions buyers ask most when comparing Vapi and Retell AI.
How much do Vapi and Retell AI cost?
Vapi charges $0.05 per minute for hosting plus transcriber, model, voice and telephony at provider rates, and its calculator puts 1,000 minutes at $82 to $129 a month. Retell AI quotes $0.07 to $0.31 per minute all-in, and a GPT 5.1 build on a Retell number comes to $0.125 per minute. Both prices were checked on 20 November 2026.
Is Vapi cheaper than Retell AI?
At 1,000 minutes a month with mid-range models and voices, the per-minute totals are close. The gap usually comes from concurrency: Retell includes 20 lines, while Vapi includes 4 on usage-only and 10 on $29 Core, with extra lines at $10 a month. Model both at your peak hour.
Do Vapi and Retell AI offer free credits?
Yes. Vapi gives new accounts $5 in free credits with no minimum minutes, and Retell gives $10 on pay as you go. Neither has a free plan beyond those credits.
Can I use my own OpenAI or ElevenLabs account?
On Vapi, yes: validate your provider key in the Integrations tab and that provider bills you directly, while Vapi keeps charging its $0.05 hosting fee. Retell supports a custom LLM over WebSocket, but its voice and infrastructure rates are billed by Retell.
Which platform is better for a contact center?
Retell's integrations page lists Five9, NICE CXone, Talkdesk, LiveVox, Amazon Connect and RingCentral, and SIP telephony carries no Retell charge. Vapi supports SIP at no transport fee, but you build transfer logic with squads and tools, so contact centers usually have less to build on Retell.
Are Vapi and Retell AI HIPAA compliant?
Vapi sells HIPAA as a $2,000 per month add-on with a BAA; it applies to the whole organization and requires HIPAA-compliant providers. Retell's pricing page describes the platform as HIPAA-ready but doesn't clearly state which plan includes BAA terms. Confirm BAA terms in writing with either vendor before handling patient data.
What happens when I hit the concurrent call limit?
On Vapi, new inbound and outbound calls are blocked until a slot frees up. On Retell, calls above the limit can connect if you enable burst pricing, which adds $0.10 per minute for the whole call. Without burst, you'd buy extra lines at $8 each a month on Retell or $10 on Vapi.