Most pricing pages in this category answer a narrower question than the one you asked. They give you one monthly number and leave out the other bills that arrive with it. Below is the whole picture, with public competitor prices as we checked them in July 2026, the fees Meta charges directly, what a human setter costs, and the arithmetic for deciding whether any of it pays for itself in your business.
What are you actually paying for?
There are four cost layers in every AI sales agent. Vendors tend to quote one and stay quiet about the rest.
Layer 1: the platform. This is the subscription: the software that receives the message, decides what to say, books into your calendar, sends the payment link, and stores the history. It is the number on the pricing page, and it is usually the only number people compare.
Layer 2: WhatsApp messages. Meta charges for certain messages, and it charges you, not your vendor. This bill lands in your Meta Business account no matter which platform sits on top. More on how it works in a moment.
Layer 3: the language model. If you buy a finished product, this is baked into the subscription. If you build your own agent, you pay per token for every message the model reads and writes, including the ones it gets wrong and retries. That bill grows with your volume rather than sitting flat.
Layer 4: human time. Somebody configures the thing, and somebody keeps it correct when your prices, hours or services change. On a flow builder this is the ongoing cost of redrawing branches. On HeySetter the guided setup is 6 steps, and updates happen through Coach Mode: when the agent does not know something, it asks you on WhatsApp and saves your answer instead of waiting for you to edit a prompt.
A fair comparison between two options adds up all four layers. Comparing layer 1 alone is how people end up surprised on month two.
Three ways this gets billed, and who each one punishes
Price is not just a number, it is a shape. The shape decides who ends up overpaying.
Per active contact. ManyChat's public prices in July 2026 run Free at $0 for 25 active contacts, Essential at $14 a month for 250, Pro at $29 for 2,500, Business at $69 for 7,500 and Advanced at $139 for 25,000, with contacts beyond your tier billed on top. The headline is cheap, and for small lists it genuinely is. It punishes anyone running paid traffic, because a contact who asks "how much" once and disappears counts exactly like a buyer, and your bill climbs with your ad spend rather than with your revenue. The automation you get is also flow-based: you draw the branches in advance and maintain them forever.
Per user or per seat. Wati's public prices in July 2026 are Growth at $29 a month billed annually ($40 month to month), Pro at $99 annually ($135 month to month) and Business at $299 annually ($399 month to month), with extra users charged separately. This model punishes teams that grow. Your cost tracks headcount, which is the opposite of what you want from automation, and the attractive numbers require an annual commitment.
Per unique conversation. HeySetter charges $350 a month with 3,000 unique conversations included and $0.08 for each one after that. A unique conversation is one contact who exchanges at least one message with your agent during the billing month, whether that is 2 messages or 200. This model punishes very high volumes of conversations that never buy: 8,000 unique conversations in a month costs $350 plus 5,000 at $0.08, which is $750. If that is your normal month, the honest answer is that Setter Premium fits better, and it is priced for your company on a call.
Building it yourself. The prototype is a weekend. The product is not. You are signing up to own WhatsApp Cloud API webhooks, retry and queue logic, calendar availability math, payment webhooks, template approvals, quiet hours, and a telemetry trail for when the agent does something you did not expect. The subscription you avoided is replaced by engineering time that never ends, plus a model bill that scales with every conversation.
What does Meta charge for WhatsApp messages?
This is the layer most comparisons get wrong, so it is worth stating precisely. Since July 1, 2025, Meta's WhatsApp Business Platform bills per delivered template message rather than per conversation. Replies sent inside an open customer service window are not charged. Rates vary by the recipient's country and by the template category: marketing, utility, authentication or service.
Two consequences matter for your budget. First, Meta bills your Meta Business account directly, separately from any platform fee, so this cost is identical whichever vendor you pick and no vendor can make it disappear. Second, your real WhatsApp spend depends on how much of your messaging happens inside the free window versus how many marketing templates you push out. An agent that answers fast, inside the window, is cheaper to run than one that reopens conversations with paid templates.
Run your break-even number, then test it for a week. 7 days, every feature, no card.
Start freeIs an AI agent cheaper than a human appointment setter?
According to ZipRecruiter in July 2026, the average appointment setter in the United States earns $50,455 a year, roughly $24.26 an hour, with the middle of the market between $33,000 and $62,000. That is about $4,200 a month on average, and roughly $2,750 to $5,170 a month across that middle band, before payroll taxes, benefits, software, training and the management time it takes to keep one person productive.
At $350 a month, the agent costs about 8% of that average monthly salary, and it answers in under 30 seconds at 2 AM, on Sundays, in English and Spanish, without forgetting the follow-up at 2 hours, 8 hours and 48 hours.
Now the honest part. A human does things this agent does not do. They pick up a phone and chase a cold list. They hear hesitation in someone's voice and change the pitch. They handle the messy negotiation, the exception, the angry customer who needs a person. The useful comparison is not replacement, it is coverage: the agent takes the repetitive volume so a human is free for the conversations that actually need judgment. If your plan is to fire your team and install software, that is not what this is.
When it is not worth it
Three situations where the math does not work, stated plainly.
Very low lead volume. If you get 10 to 20 WhatsApp inquiries a month and you personally answer all of them within minutes, $350 buys you almost nothing you do not already have. Speed is the main lever here, and you already pull it.
Nothing to book and nothing to charge. Two of the strongest parts of the product are booking into your calendar and collecting the deposit through Stripe or Bold. If your sale ends with someone walking into a store, with no appointment and no prepayment, you are paying for a stack you will use half of.
A ticket too small to carry the cost. If your average sale leaves you $15 of profit, the agent needs to produce about 24 additional paid sales every month just to reach zero. That is possible in high-volume ecommerce and unrealistic almost everywhere else. Run the number below before you decide.
How to calculate your own break-even
The formula fits in one sentence: take the monthly cost of the agent and divide it by the profit you keep from one appointment that actually gets paid. The result is how many extra paid appointments per month the agent has to generate before it costs you nothing. Everything above that line is yours.
If you expect to run past 3,000 unique conversations, add $0.08 for each extra conversation to the monthly cost before dividing.
Simulated example: a clinic charges $250 for a first treatment and keeps $100 after materials and the practitioner's cut. Divide $350 by $100 and you get 3.5, so 4 extra paid appointments a month puts the clinic ahead. The clinic receives about 300 WhatsApp inquiries a month and currently replies to many of them the next day. Turning 4 of those 300 into paid bookings is a lift of a little over 1%. That is the question worth asking: not whether the software is impressive, but whether a 1% lift is plausible in your business.
Run your own version of that number. If it comes out obviously reachable, the 7-day trial costs you nothing but a week of attention. If it comes out uncomfortable, you just saved yourself $350.
Frequently asked questions
None of the three. The plan is $350 a month, billed monthly, cancelled from the dashboard whenever you want, with access until the end of the period you already paid for. There is no implementation charge, no permanence clause and no cancellation penalty.
One contact who exchanges at least one message with your agent inside the billing month, counted once no matter how many messages follow. The counter resets each calendar month. Past 3,000, each additional conversation is added to your next invoice at $0.08.
The AI, the platform and the infrastructure are included, with no separate token bill. WhatsApp message fees are not, because Meta charges them straight to your own Meta Business account under its own country and category rates. We show you where to read those numbers during setup.
It is priced for each company on a call. The figure depends on your volume, the integrations you need and the custom work involved, so there is no public number and no starting price. Book a call and you leave with a real quote rather than an estimate.
Yes. The trial runs 7 days with full access to every feature and asks for no card. If you decide against it, nothing renews and nothing is charged. Most people use the week to watch the agent handle real inbound messages and judge the quality themselves.
No. The plan is $350 USD everywhere, and prices are always quoted in US dollars. What does vary by country is Meta's WhatsApp message rate, which is billed to you directly and depends on where your customers are and which template category you send.
