Blog · July 25, 2026 · 8 min read

Retell AI Pricing Explained: What It Costs to Run Voice Agents (and How Agencies Make Margin)

By Nabeel Hassan — builder of VoiceDash

Retell AI Pricing Explained: What It Costs to Run Voice Agents (and How Agencies Make Margin)

TL;DR: Retell AI charges per minute of conversation, billed pay-as-you-go, and that per-minute rate is only part of your true cost. The full number is the voice engine, plus telephony (usually Twilio), plus any LLM and text-to-speech pass-through, plus concurrency you provision for busy hours. I build production voice agents on Retell AI for US clients and run VoiceDash, a white-label client portal for voice AI agencies, so here is how Retell pricing actually works, what a real call costs you, and how agencies turn a few cents a minute into a healthy retainer.

Retell publishes a per-minute number and most people stop reading there. Then their first invoice surprises them, because the sticker rate is the engine, not the whole car. If you are pricing a client, quoting a retainer, or just trying to understand your margins, you need the total cost of a minute, not the headline. Let me walk through it the way I model it before I quote anyone.

How Retell AI prices: per minute, pay as you go

Retell bills by the minute of active conversation, on a usage basis, with volume discounts as your monthly minutes climb. There is no seat license and no per-agent fee in the way a SaaS tool charges. You pay for the minutes your agents actually talk.

At the time of writing the all-in voice engine lands somewhere in the range of a few cents to roughly a dime per minute depending on your volume tier and the models you pick, but Retell adjusts pricing and tiers regularly, so treat any number you read online, including this one, as a starting estimate and confirm the current rate on Retell's own pricing page before you quote a client. The structure matters more than the exact cents, because the structure is what stays true.

The important mental model: Retell's per-minute rate covers the orchestration of the voice pipeline, speech-to-text, the LLM turn, and text-to-speech, stitched together with low latency. That is the piece Retell is genuinely selling you. Everything around it is a separate line you also have to pay.

The costs people forget to add

Here is where the real invoice comes from, and where new agency owners get caught.

Telephony (Twilio)

Retell runs the conversation, but the phone number and the carrier connection are usually Twilio. Twilio charges you separately for the number rental each month and a per-minute rate for inbound and outbound calls. It is small per minute, but it is real, and it stacks on top of Retell's engine minute. When I estimate a call's cost, I always add the Twilio leg. I break down where Twilio fits in the wider stack in how to start a voice AI agency.

LLM and voice model pass-through

Depending on your configuration, the language model and premium voices can be billed as pass-through or bundled, and the model you choose changes the number. A cheaper model shaves cost per minute; a premium, more natural voice costs more. For an inbound receptionist that has to sound human, I spend the extra on voice quality, because a robotic agent loses the client the trust that justifies the whole retainer.

Concurrency for busy hours

A dental office does not get calls evenly across the day. It gets a rush at open, a spike at lunch, and a wave at close. To answer every call during those spikes you provision for concurrent calls, and provisioning is a cost decision, not a free default. Under-provision and you drop calls during exactly the moments that matter, which is the failure the client hired you to prevent.

What a single call actually costs

Put it together with round numbers so the shape is clear. Say a typical receptionist call runs three minutes. You pay Retell's engine rate for those three minutes, plus Twilio's per-minute leg for three minutes, plus the number rental spread across all your calls, plus whatever model and voice premium your config carries. In practice a three-minute inbound call costs an agency somewhere in the low tens of cents, not dollars.

That is the whole point. Your raw cost to answer a call is small. The value you deliver, a booked appointment or a captured lead that would otherwise have gone to voicemail and then to the competitor on Google, is worth far more than the pennies the call cost. The gap between those two numbers is your margin, and it is enormous when you price it right.

How agencies make margin on top of Retell

This is the part that actually matters for your business. Retell is your cost of goods. Your job is to price the outcome, not the minute.

The mistake I see constantly is agencies passing Retell's pricing model straight through to the client, billing per minute with a markup. It feels fair and it quietly kills your margins, because it caps your upside, turns your client into an accountant optimizing your line item, and exposes your markup the moment they glance at a Retell invoice. I made the full case against per-minute billing in voice AI agency pricing.

What works instead:

  • Charge a flat monthly retainer. A booking is worth the same to the client whether the call took ninety seconds or four minutes, so bill on the outcome and the peace of mind, not the meter. Most healthy receptionist retainers land in the hundreds to low thousands a month while your underlying Retell and Twilio cost is a fraction of that.
  • Meter minutes privately. You still need to watch usage, because a client whose call volume triples is eating into your margin. Track it in your own view, and if someone genuinely runs hot, that is a conversation about a higher tier, not a per-minute invoice.
  • Add a setup fee. The build, the prompt, the number provisioning, and the go-live are real work you do once. Charge for it up front so you are not financing the client's onboarding out of month one.

The economics are excellent precisely because Retell's per-minute cost is low and predictable. You are buying pennies and selling an answered phone.

Why usage visibility protects your margin

Here is the operational catch. Retell's cost model is easy to lose track of across several clients, because minutes accumulate quietly in the background and you only notice on the invoice. If one client's volume climbs, your margin on that account erodes and you may not see it until the month closes.

That is one reason usage metering sits inside the product I build. When I run a client on VoiceDash, I can see minutes and call volume per client in one branded place instead of cross-referencing raw Retell exports. The client sees their own calls, outcomes, and recordings in a portal on my domain; I see the usage that tells me whether an account is still profitable. It is the same data serving two jobs, which is exactly what I argued reporting should do in voice AI client reporting.

Visibility is not a nice-to-have here. It is how you keep a flat retainer profitable while a client's call volume grows, and how you know when it is time to move someone to a bigger package before the margin disappears.

Does the platform's pricing even matter to your client?

Almost not at all, and this is worth internalizing. Clients never ask what Retell costs per minute. They ask whether the phone gets answered and whether they can see it working. Your platform pricing shapes your margins and your quoting; the client experience comes from the branded layer you put on top, which is the same conclusion I reached comparing engines in Retell AI vs VAPI vs Bland.

So understand Retell's pricing cold, because it decides your profit. Then never make it the client's problem. Sell them an outcome, deliver it through a portal that looks like software you built, and let the pennies-per-minute economics stay quietly on your side of the ledger.

The pricing math in one line

Retell charges per minute, telephony and models add to it, and a three-minute call costs you cents. You resell that as a flat retainer worth hundreds or thousands, meter usage privately so a heavy account never silently eats your margin, and wrap the whole thing in a branded client portal so the client sees a product, not a meter. Get that stack right and voice AI is one of the highest-margin services you can sell.

Want to keep every client's minutes and outcomes visible in one branded place so your margins stay healthy as you grow? Start free on VoiceDash or book a demo and I will show you how a white-label portal turns Retell's low per-minute cost into a retainer that renews.

FAQ

How much does Retell AI cost?

Retell AI bills per minute of active conversation, pay-as-you-go, with volume discounts as your monthly minutes rise, and no per-seat or per-agent license. The published per-minute rate covers the voice engine that stitches speech-to-text, the LLM turn, and text-to-speech together with low latency. Your true cost is higher than that headline, though, because you also pay for telephony (usually a Twilio number rental plus a per-minute leg), any LLM and premium voice pass-through, and concurrency you provision for busy hours. Retell changes its tiers regularly, so confirm the current rate on Retell's pricing page before you quote a client.

What does a single Retell AI call actually cost an agency?

Less than most people expect. A typical three-minute inbound receptionist call costs an agency somewhere in the low tens of cents once you add Retell's engine minutes, Twilio's per-minute leg, the number rental spread across all calls, and your model and voice choice. The raw cost to answer a call is small, while the value, a booked appointment or a captured lead that would otherwise have gone to voicemail, is worth far more. That gap between pennies of cost and a real business outcome is exactly where your margin lives.

Should I charge clients per minute like Retell charges me?

No. Passing Retell's per-minute model straight through with a markup caps your upside, turns the client into an accountant optimizing your line item, and exposes your markup the moment they see a Retell invoice. Charge a flat monthly retainer priced on the outcome instead, add a setup fee for the build, and meter minutes privately so you can spot an account whose volume is eroding your margin. A white-label portal like VoiceDash shows the client their own outcomes while giving you the per-client usage visibility that keeps a flat retainer profitable as you grow.

Give your clients a dashboard with your name on it

VoiceDash turns your Retell agents into branded client portals. Live in under 10 minutes, no code.

Keep reading