How much does a WhatsApp chatbot cost in Mexico?

The entry price each vendor publishes, in the currency it publishes it in, and where the AI agent sits. Each one has its full pricing guide at the end of this page.

This is the advertised number, and it runs from $32 USD per month on Chatbase's entry plan to $247 USD per month on Leadsales' Avanzado plan. In between: ManyChat Pro at $39 USD, Whaticket's entry plan at $49 USD with three users and one WhatsApp connection, Wizybot from $69.99 USD and Leadsales Profesional at $133 USD; Chatbase reaches $400 USD per month depending on credits and seats.

Kommo advertises the lowest entry figure of all, $20 USD, but it is per user per month and for new customers (customers who bought before 1 September 2026 keep fifteen dollars while they stay on their plan), and it sells no monthly option: the published minimum is six months prepaid, meaning $120 USD per seat before the first conversation. In pesos, Aurora Inbox publishes $1,800 MXN per month plus VAT with three users included.

Eva, our product, starts at $191 USD per month on an annual plan ($229 USD month to month) on the Essential plan, with 5,000 credits and 5 users included, and artificial intelligence already inside the price.

Before comparing two subscriptions, check what each one carries: how many users, how many channels, how many WhatsApp connections, and how the entry plan connects WhatsApp. Leadsales' Plan Básico, at $97 USD per month, connects it through WhatsApp Web with a QR code; the official WhatsApp API starts on Profesional.

This is where two products with the same subscription end up costing twice as much as each other. On several platforms the subscription buys the inbox and the flows, and artificial intelligence is a separate line.

Leadsales bills its agent apart: Lead Agent costs $87 USD per month on Starter, $297 USD on Growth and $793 USD on Scale. Kommo includes credits per user and sells extra packages at $29 USD and $99 USD per month, with at least three months prepaid on a first purchase. Wizybot includes one thousand two hundred and fifty artificial-intelligence messages on its entry plan and charges $0.07 USD per extra message. Chatbase charges $40 USD per thousand credits on top-up.

Whaticket is the extreme case on the other side: its plan does not include the artificial-intelligence layer at all, and its usage is paid straight to OpenAI with no published allowance or per-conversation consumption, so that budget line stays blank until you measure it yourself. With Eva the credits belong to the plan and a 1,000-credit top-up costs $49 USD.

The arithmetic that helps: take one real month of conversations, multiply by the replies each one takes, and convert that into whatever unit the vendor bills. If the result exceeds what is included, the true price is the subscription plus the top-ups, not the subscription.

The WhatsApp API belongs to Meta, not to the platform you hire. Since 1 July 2025 Meta charges separately for each delivered template message, by template category, and none of the platforms in this guide folds that fee into its subscription: some resell it as balance you top up, others ask you to pay Meta directly.

The model has two parts worth understanding before budgeting. The rate changes by country and by message category: a marketing template does not cost what a utility or authentication template costs. And there is a 24-hour service window the customer opens by writing first. From October 1, 2026, under Meta's official rate card, utility templates delivered inside that window are charged, and service messages are free up to 1,000 delivered per phone number per month.

On Meta's official rate card in force from October 1, 2026, a marketing template delivered in Mexico costs USD 0.0397 and a utility or authentication template USD 0.0085. The detail, with the free allowance of service messages, is in the WhatsApp Business API pricing guide. Ask your vendor in writing whether it resells those fees with a margin or passes them through at cost.

An agent is useful in proportion to how much it knows about your business. Loading the catalog with the fields people choose on, writing the price and shipping terms, defining appointment types with real durations, connecting the number and testing genuine conversations is work, and somebody does it: you, your team or the vendor.

Vendors solve this differently. Kosmo includes it in all three plans: they analyze the business, configure it, connect the channels and train the team, with a published 72-hour turnaround. Its entry plan costs $4,497 MXN per month. Setup is included: we configure everything Eva does today, at no extra cost.

If somebody quotes you a custom build, ask for the breakdown in writing and by concept: discovery, configuration, integration with your system, testing, and who maintains the thing when your catalog changes. A closed price without that breakdown almost always turns into a change of scope in month two.

The first is per-person multiplication. When pricing is per user, your cost is decided by your headcount rather than by the tool: on Kommo, going from three people to eight more than doubles the bill, and all users on the account share the same plan, so levels cannot be mixed seat by seat.

The second is that credits expire. On several platforms the included allowance resets monthly and does not roll over: a slow month does not finance the next one, and a busy month is paid in top-ups. Check what happens at the ceiling too, since some switch artificial intelligence off until the next cycle instead of overcharging you.

The third is the minimum term, which turns a trial into a commitment. And the fourth is the invoice: most bill in dollars, with no Mexican entity or local tax invoicing, so the real cost includes the day's exchange rate and the accounting work of a foreign receipt.

There is a fifth that shows up once the thing is working: plan extras. Chatbase, for instance, charges $1,188 USD a year to remove its branding from the chat window, and on its free plan agents are deleted after fourteen days without activity. Neither line appears in the first comparison, and both change the decision.

If what you get is five repeated questions, such as opening hours, address, parking, directions or whether you are open today, a button flow on the cheapest plan on the market handles them and you need nothing else. Putting an artificial-intelligence agent there is paying for capacity you will not use.

The line is crossed when the answer depends on data that moves. If something has to be looked up in a catalog with stock and prices, quoted with terms, checked against real calendar availability, charged for, or understood from a photo, a voice note or a question nobody anticipated, the button flow breaks and the conversation ends in «let me connect you with an advisor», which is exactly where you were before paying.

Volume crosses it too. A cheap flow answering two hundred people badly each month costs more than an agent answering them well, because the real cost is not the subscription: it is the sales that fall through and the hours your team spends fixing what the flow left half done.

The honest way to decide: read fifty real conversations from last week and count how many are settled with a fixed answer. If nearly all of them are, buy cheap. If half of them require searching, quoting or booking, you already know what you are actually buying.