Here’s the short version. Synthflow is a no-code voice-AI builder, so you drag boxes around and get an agent live without writing a line. Retell and Vapi are developer platforms, so you build the agent in code or over an API and get far more control over calls, functions, and telephony. On price, Synthflow lands around $0.16 a minute, Retell blends to $0.07 to $0.31, and Vapi starts at $0.05. The catch with no-code isn’t the price. It’s the ceiling: custom integrations, telephony control, a signed BAA, and complex call routing are where a drag-and-drop builder runs out of room and pushes serious buyers toward code or a done-for-you build. Reach out and we’ll tell you which layer your call actually needs.
We’re gmware, a software development firm headquartered in Austin, TX with engineering centers in Bangalore and Mohali, India, and AI agent builds are one of our core service lines. Our angle here is simple: we build and operate the agent for you on whatever layer fits, no-code where it’s enough and custom where it isn’t. Below is the comparison table, the per-minute math, the five things that push a buyer off no-code, and an honest call on when a no-code builder is all you need.
| Synthflow (no-code) | Retell (dev platform) | Vapi (dev platform) | Done-for-you build | |
|---|---|---|---|---|
| Who codes it | You, in a visual builder | Your developers | Your developers | A team you hire (us) |
| Integration depth | Prebuilt connectors, ceiling on custom | Deep, code-level | Deep, code-level | Whatever your systems need |
| Telephony control | Limited, managed for you | High (SIP, transfers) | High (SIP, transfers) | Full, designed to your routing |
| Compliance / BAA path | Constrained | Design-your-own | HIPAA add-on $2,000/mo | Built and signed for you |
| Pricing model | Usage, ~$0.16/min | Stacked, $0.07 to $0.31/min | From $0.05/min | Scoped per engagement |
| Best-fit buyer | Simple booking / FAQ bot | Devs wanting control | Devs wanting control + BYO | Regulated / complex / hands-off |
Read that table left to right and you can see the trade. Synthflow gives you speed and hands off the plumbing. The two platforms give you control and hand you the wiring. A done-for-you build gives you both, minus the part where you own any of it.
Per-minute cost, at a glance
What is the actual difference between Synthflow, Retell, and Vapi?
The difference is who does the building and how far down you can reach. Synthflow is no-code: a visual builder where you assemble an agent from prebuilt blocks, connect a calendar or a CRM through a supported connector, and publish. No engineer required. That’s the whole pitch, and it’s a good one for the right job.
Retell and Vapi are developer platforms. You build the agent in code or through an API, define your own functions, control how calls connect and transfer, and pick your own models. Retell bundles speech-to-text, its voice infrastructure, and text-to-speech, then lets you plug in the LLM and telephony you want. Vapi charges a flat voice rate and lets you bring your own model keys, so model cost can drop to zero on your account. Both assume someone on your side writes code.
So the split isn’t feature-for-feature. It’s a control-versus-effort trade. No-code hands you a working agent fast and caps how deep you can customize. A dev platform hands you the depth and asks for engineering time. Neither is wrong. They’re built for different buyers.
One quick note on scope. This post is the no-code-vs-platform-vs-build question. If you’re already committed to writing code and just want to compare the developer SDKs head to head, we broke Vapi vs Retell vs Bland AI down separately. That one’s the dev-tool cage match. This one’s the layer decision.
How much do Synthflow, Retell, and Vapi cost per minute?
Every one of these bills by usage, and every published rate is a floor, not a ceiling. Here’s what each vendor actually charges.
Synthflow moved to pay-as-you-go, component-based billing: a voice engine cost, an LLM cost, and a telephony cost, stacked. A common config lands around $0.16 a minute, and most setups sit somewhere between $0.11 and $0.24 depending on which LLM and telephony you pick. Big contracts go the other way: Synthflow enterprise starts at $30,000 a year.
Retell stacks its own line items: infrastructure at $0.055 a minute, text-to-speech at $0.015 a minute, the LLM (varies by model), and telephony at roughly $0.015 a minute. Speech-to-text, infra, and TTS come bundled. That blends out to $0.07 to $0.31 a minute across the range. Phone numbers run $2 a month, concurrency is free up to 20 lines then $8 a month beyond that, and new accounts get $10 in free credits.
Vapi is the cheapest sticker: $0.05 a minute for voice on the Build plan, with model costs charged at cost or $0 if you bring your own key. You get 10 concurrent lines free, then $10 per line per month. The compliance add-ons are separate and not cheap: $2,000 a month for HIPAA and $1,000 a month for zero data retention.
Three ways to stack a per-minute bill
Here’s the honest read on price. At low volume the difference between $0.05 and $0.16 a minute is real money but small dollars, so it shouldn’t drive the decision. What drives it is control and compliance. Vapi’s HIPAA add-on alone is $24,000 a year. That single line reorders the whole comparison for a regulated buyer, and it’s why per-minute rate is the last thing you should optimize. If a BAA and real integrations are on your list, reach out and we’ll tell you whether no-code even clears your bar.
What pushes a buyer off a no-code voice builder?
No-code is genuinely great until it isn’t, and the wall shows up in five predictable places. We’ve watched buyers pick Synthflow, build a working agent in a day, then hit one of these three months later and start over on a platform. Better to see the wall before you walk into it.
The no-code ceiling: 5 things that push you off
The first is a custom integration. If the agent has to talk to a system no connector supports, an old scheduling tool, a homegrown CRM, an internal API behind auth, no-code either can’t do it or forces an awkward middle layer. On a dev platform you just write the function.
The second is telephony. Managed telephony is fine for a straight inbound line. The moment you need SIP trunks, warm transfers to a specific human, or call routing that changes by time of day or caller, you want the control a platform gives you. A no-code builder decides that for you.
Third is compliance. If a call carries PHI or regulated data, you need a signed BAA and strict data-retention terms. That’s exactly why Vapi sells a $2,000-a-month HIPAA add-on: compliance is a design decision, not a checkbox. No-code limits how you route and store data, which is where regulated buyers stall.
Fourth is complex routing. A booking bot with one path is easy. An agent that qualifies a lead, checks three conditions, and branches to different actions is a state machine, and state machines are code. Fifth is volume. At a few thousand minutes a month, per-minute rate barely matters. At high volume it compounds, and owning your own model keys (Vapi’s BYO) or a tuned build starts to pay for itself.
When is a no-code builder like Synthflow all you need?
Sometimes it is, and we’ll say so on a scoping call before we’d sell you anything heavier. Here’s an opinion we’ll defend: for a lot of small businesses, a no-code voice builder is the right answer and a custom build would be a waste of money.
Picture a plumber who misses every call after 6pm. He needs one thing: an agent that answers, takes the address and the problem, books a slot on his one calendar, and texts him. One integration. One path. No compliance. No routing rules. Synthflow does that in an afternoon for a few dollars a day of usage, and nothing about a $40,000 build would serve him better. If that’s your shape, use the no-code tool. We mean it.
The fit signals are clear. Your calls follow one or two simple paths. You integrate with one common system a connector already supports. No PHI, no BAA, no data-residency rule. Modest call volume. You want it live this week, not next quarter. Check those boxes and no-code is the cheapest correct answer, full stop.
The build signals are the mirror image: an integration no connector covers, telephony you need to control, a BAA to sign, routing that branches, or volume where owning the stack pays off. That’s the same buy-versus-build line we walk in what an AI receptionist costs and in the broader case for AI agents across business operations. The tool isn’t the decision. Your requirements are.
How gmware builds and operates a voice agent for you
We don’t sell you a layer. We pick the one your call volume, integrations, and compliance actually need, then build and run the agent on it so you don’t have to learn Synthflow, Retell, or Vapi at all.
That means no-code where no-code is genuinely enough, and we’ll tell you when it is even though there’s less in it for us. It means a developer platform like Retell or Vapi when you need custom functions, telephony control, or a compliance path we design and sign for you. And it means a fully custom build when your systems or your routing outgrow every off-the-shelf option. We run production data systems ourselves, Shield Suite tracks beverage-alcohol brands across 60,000+ storefronts, so the reliability and guardrails part isn’t theory to us.
If you’re weighing this against a human answering service instead of another AI tool, the price math lives in our virtual receptionist pricing breakdown, and the wider integration cost picture is in what it costs to add AI to existing software.
Tell us what your calls look like: the volume, the systems, and whether a BAA is in the picture. Reach out and we’ll give you a straight answer on which layer fits, what it costs, and how long it takes, within 48 hours. Sometimes that answer is “just use Synthflow,” and we’ll say that too.