So you have narrowed your shortlist to two names that come up again and again when people talk about building an AI phone agent: Vapi and Retell. Good choice to put them head to head, because on the surface they look almost identical. Both run the call. Both let an AI answer the phone, talk to a caller, book a slot, qualify a lead or take a payment. Both charge by the minute. The difference is in how they assemble the call and what they hand you on the bill, and that difference is the whole decision.
Quick map of where this goes. First the honest version of the price, because the headline number on both sites is misleading in different ways. Then who each one is actually built for. Then where each one genuinely wins, with the names of real companies using them. Then a worked cost example, integrations, compliance, the bit we have not measured yet, and a straight answer at the end.
The price, told honestly
Vapi charges $0.05 a minute to host the call, and that is the only number Vapi sets. The three moving parts of any voice agent, turning speech into text, the AI working out a reply, and turning that reply back into a voice, are billed straight through from whoever you plug in, at their rates, with no Vapi markup when you bring your own API keys. The phone line comes from your carrier. So the floor is genuinely cheap, and your real number is whatever your chosen parts add on top. In practice that lands somewhere between $0.05 and $0.30 a minute once a normal speech, model and voice stack is wired in.
Retell does it the other way. Its engine rate is around $0.07 a minute, and that rate already folds in most of the real-time voice pipeline. You still add a language model, a voice and a phone line on top, so the all-in number is not just $0.07, but more of the plumbing is bundled into the one rate. In practice Retell lands between $0.13 and $0.31 a minute all-in.
Read what that means rather than the raw figures. Vapi’s floor is lower and its ceiling is about the same, so if you are willing to shop for cheaper components and run your own keys, Vapi can come out cheaper per minute. Retell starts higher because you are paying for the bundling. You are buying fewer decisions. That is not a worse deal, it is a different deal, and which one is better depends entirely on whether you want to make those decisions.
One number people miss: concurrency, the cap on how many calls run at once. Vapi includes 10 lines, then charges $10 per extra line a month. Retell includes 20, then charges $8 each. If you are running an outbound campaign that dials hundreds of numbers in an hour, that included headroom and the per-line price matter as much as the per-minute rate, so do the sum for your own call pattern before you decide either is cheaper.
Who each one is built for
Here is where I want to correct a common framing. People describe this match-up as “developer-first Vapi versus no-code Retell”. That is not quite right, and I checked. Both are API-first. Both expect you, or someone on your team, to do real configuration. The honest distinction is narrower than the marketing suggests.
Vapi is unapologetically a developer’s tool. Its own pitch is “API-first by design”, and the whole product is built around the idea that you assemble the pieces. Want to swap one speech-to-text provider for another, or run a cheaper model on the easy questions and a smarter one on the hard ones? You can, and every choice shows up on the bill and in the latency. If you have an engineer, that control is the point. If you do not, it will feel like more wiring than you wanted.
Retell leans the same API-first way but wraps more around it. It frames itself around the contact centre, “supercharge your contact-centre operations”, and ships a fuller dashboard for building and watching agents. So a less technical operations lead can get further inside Retell before they hit a wall that needs a developer. Both still reward a technical owner. Retell just lowers the first step onto the ladder.
That gives you the first two use-case fits, and they are clean:
- You are a product or engineering team that wants control over cost and behaviour. Vapi. You will use the flexibility, and the flexibility is most of what you are paying that low platform fee for.
- You are a contact centre or an ops team that wants to ship without standing up a component stack. Retell. The bundled pipeline and the dashboard are doing real work for you, and the higher floor buys back time.
Where Vapi wins
Vapi’s strongest card is scale and the trust that comes with it. As of May 2026 it had raised at a reported $500M valuation, and the customer list is the kind that does serious due diligence. Amazon Ring, by TechCrunch’s account, routes all of its inbound calls through Vapi after evaluating more than forty rival platforms. Intuit is a named customer too. Those are not logos a finance team picks lightly, and for a buyer worried about whether a young category is safe to build a phone line on, that track record is reassuring.
The second win is the flexibility we have already covered, turned into money. Because you can bring your own keys and choose cheaper components, a team that is willing to tune can run Vapi near the bottom of that price range. Retell does not give you that lever in the same way, because the lever is exactly what it bundled away.
The third is the operational kit. Vapi carries SIP trunking, so you can plug in your own phone-number supplier instead of using its numbers. It does warm transfers, handing a live call to a human with the AI’s summary attached. It runs outbound campaigns in bulk. And it speaks MCP, the connection that lets other AI tools trigger and feed calls. Retell matches most of this, with one gap noted below.
Where Retell wins
Retell’s win is the flip side of the same coin: it asks less of you. The bundled engine means fewer providers to choose, fewer keys to manage, fewer bills to reconcile, and a dashboard that shows you what is happening without a console. For a team whose scarce resource is engineering time, not money, that is worth the higher per-minute floor. You are buying back the hours you would otherwise spend assembling and maintaining the stack.
It has real customers to point to as well. Retell features Everise, a large customer-experience operator, and GiftHealth among its case studies, and says more than 3,000 businesses use it. A caution worth stating plainly: the headline outcomes on those case studies, the “cut costs by half” sort of number, are Retell’s own reported figures, not anything we have independently checked, so read them as the vendor’s claim rather than as proof.
On capabilities, Retell carries SIP trunking, warm transfer and batch calling, the same core call kit as Vapi. The one gap: it does not support MCP, where Vapi does. If letting other AI tools orchestrate your calls is on your roadmap, that is a real difference. If it is not, you will never notice it.
A worked example, so the numbers feel real
Say you run 1,000 minutes of calls a month, a small but real outbound or support load. On Vapi the platform fee alone is about $50 for those minutes, and then your speech, model, voice and phone line stack on top; depending on the components you choose, the all-in cost lands roughly between $50 and $300 for the month. On Retell the same 1,000 minutes start from a higher engine rate and land roughly between $130 and $310 all-in once the model, voice and line are added.
The point of that range is not to hand you a single figure, because an honest single figure does not exist until you pick your components. The point is the shape: at the cheap end, a tuned Vapi setup is clearly the lower bill, and at the busy, fully-featured end the two converge. So if cost is your first constraint and you are willing to optimise, Vapi has the edge. If you would rather not think about which model is running, the premium for Retell is the price of not thinking about it. Run your own minutes through the cost calculator before you commit, because your call mix, not our range, decides the real number.
Integrations and the rest of your stack
A voice agent rarely lives alone. It needs to read from and write to a CRM, hand off to a human, and fit whatever phone setup you already run.
Retell publishes a clear list of connectors on its batch-calling product, including HubSpot, GoHighLevel, Twilio, Vonage and n8n, which makes it easy to wire into a sales or support workflow. One honest caveat from Retell’s own writing: for CRMs it does not connect to directly, the documented path is exporting your contacts and uploading them, rather than a live two-way sync. So check that your specific CRM is a real integration, not a spreadsheet round-trip.
Vapi’s CRM story is looser in public. Its marketing describes logging qualified leads into systems like HubSpot and Salesforce, but that framing sits on a blog rather than a product integrations page, so treat it as a direction of travel rather than a guarantee, and confirm the exact connector you need. On warm transfer there is a specific limit worth knowing: Vapi’s documented warm-transfer flow is tied to Twilio-based telephony, so if you run a different carrier, test that path early.
Compliance and trust
If you are in healthcare, finance or anywhere regulated, this section may decide it for you, so here are the specifics.
Vapi offers HIPAA, but as a paid add-on at $2,000 a month, and switching it on means no logs, recordings or transcripts are kept. Zero Data Retention, which keeps nothing at all, is a separate $1,000 a month. SOC 2 Type II, GDPR and PCI DSS are covered at the platform layer, though SOC 2 sits on the enterprise plan. One caveat for buyers: Vapi only runs the call, so end-to-end coverage also leans on your phone-line and storage suppliers holding their own certifications.
Retell presents SOC 2 certification and HIPAA with a business associate agreement available, and it does not publish an equivalent flat monthly HIPAA surcharge the way Vapi does. That does not automatically make it cheaper for a compliant setup, because the all-in numbers still differ, but it does make the compliance path look less like a bolt-on. If a signed BAA and SOC 2 are gating requirements, both can clear the bar; the difference is how the cost is packaged.
What we have not tested yet
Time for the honest limit. Vapi advertises “sub-500ms latency”, and low latency is the thing that makes a voice agent feel human rather than awkward. But that is Vapi’s own marketing claim, not a number we measured. We have not placed our own timed test calls to either platform yet, so you will not find a Voxrater latency figure for Vapi or Retell on this page. When the test rig ships, we will run the same outbound scenarios against both and publish p50, p95 and the dates, and if the measured numbers contradict the marketing, the measured numbers win. Until then, treat any latency claim on either site as a claim.
The 1 to 10 scores in the table above are an editorial preview too, our provisional read from the public information, not yet from blind listening tests. They put Vapi a little ahead overall, mostly on flexibility and value, with Retell close behind on a cleaner out-of-the-box experience. Close is the honest picture here.
How locked in are you, really?
A fair worry with either platform is what happens if you want to leave. The good news is that the part that is genuinely yours, the prompt, the call flow, the logic of what your agent says and does, is portable thinking, not proprietary code, so moving the design to another platform is mostly re-entering it rather than rebuilding from nothing.
The stickier parts differ. With Vapi, because you bring your own speech, model and voice providers, those relationships are already yours; you keep your own keys and accounts, so leaving Vapi means replacing the thin hosting layer, not your whole stack. That is a real advantage of the assemble-it-yourself approach. With Retell, more of the pipeline is Retell’s, so a move means rebuilding the bundled pieces elsewhere, and any phone numbers you provisioned through the platform need porting. Neither is a trap, but Vapi’s model leaves you holding more of your own stack, which is worth a moment’s thought if you are the cautious type who likes an exit before signing the entrance. One practical tip either way: keep your prompts, call flows and test scripts in your own repository from day one, not just in the platform’s dashboard, so the design you have refined is never trapped behind a login you might one day cancel. That habit costs nothing now and saves a weekend later.
Three questions that actually decide it
If you want to skip the prose, answer these.
- Do you have a developer who will own this? Yes leans Vapi, because the control only pays off if someone uses it. No leans Retell, whose dashboard gets a non-engineer further.
- Is your scarcer resource money or time? Money leans Vapi, which you can tune to a lower bill. Time leans Retell, which trades a higher floor for far less setup.
- Do you need other AI tools to orchestrate calls through MCP? Yes means Vapi, which supports it. No, and that difference disappears.
Bottom line
Pick Vapi if you have a developer, you want to control what each call costs and how it behaves, and you are happy assembling the speech, model and voice yourself. You get the lowest floor on price, the strongest scale story in the category, MCP support, and the room to tune. The cost is your time and a fiddlier setup.
Pick Retell if you run a contact centre or a lean ops team, you would rather a mostly bundled rate and a dashboard than a pile of API keys, and you value shipping this month over shaving cents off the minute. You pay a higher floor and give up MCP and the very cheapest configurations, and in return you give up far fewer evenings.
If you are genuinely on the fence, weigh your scarcest resource one more time. Short on engineers, long on budget, go Retell. Short on budget, comfortable in a console, go Vapi. Then read the full Vapi review and Retell review for the per-plan detail, and run your own numbers in the cost calculator with your real call volume before you sign anything.