It reads intent, not spelling
Typos, slang, another language, a question wrapped in a paragraph of small talk — it still lands on the same answer, because it's matching meaning rather than a string you had to predict in advance.
Chatbot
A WhatsApp chatbot that waits before it speaks
It understands what the customer actually meant, answers from facts you wrote down, and then — before anything goes out — it pauses where you can see it. Start typing and it's cancelled. Runs inside the WhatsApp Web you already use.
7-day free trial · no credit card · works with your existing number
Why a chatbot at all
The same question, forty different ways
“Do you have a minimum order?” · “what's the MOQ?” · “qual o pedido mínimo?” · “how many do I have to buy?” — one question. A keyword rule needs all four spellings. A chatbot needs the answer.
Answers come from your text box
You write the knowledge base — prices, MOQ, shipping, opening hours, the policy you explain daily. It answers from that. A number in a reply traces back to something you wrote, not to something it made up.
Replies in their language
A question in Portuguese gets an answer in Portuguese, from a knowledge base you wrote in yours. The 2am message from a timezone you're asleep in doesn't have to wait for you to wake up.
Two modes
You decide how much rope it gets
Most “WhatsApp chatbot” tools have one setting: it's on, and it talks. Wanova has two, and you pick per rule.
Still thinking about it — what's your best price?
I can hold $4.80/unit at 500 pcs through Friday.
You always press send
Send
Draft mode — it writes, you send
The reply lands in your composer and stops there. Edit it, delete it, ignore it. Nothing reaches the customer until you decide it should. This is the default, and for anything that involves a price or a promise, it's the one we'd use.
Nothing sends without youHey, do you have a minimum order?
Yes — 500 pcs minimum, and we ship to Brazil in about 3 weeks.
holding…
Auto mode — it answers, you can veto
For the questions whose answer never changes, let it reply on its own. But it doesn't fire instantly: the reply sits visibly for a few seconds first. Start typing in that chat and it's cancelled — you take over mid-sentence, no fighting the bot.
Grace window — typing cancels itThe grace window is deliberate: Wanova never tries to guess whether you're “available” — no presence heuristics, no away-detection. It just gives you a window you can always grab.
How it works
Three steps, inside WhatsApp Web
Write what you know
One free-text box. Paste your price list, your MOQ, your shipping rules, your opening hours — the answers you already type every day. No flow builder, no decision tree, no boxes and arrows.
Rehearse on yourself
Open the test chat and play the customer. Ask it the awkward version of the question. Watch what it says back, over as many turns as you want. It touches no real conversation and sends nothing.
Pick its leash
Draft mode for anything with a price in it, auto mode for the questions you're tired of. Change your mind whenever — it's a toggle, not a migration.
Read this part
What a WhatsApp chatbot can't do for you
Everyone in this category sells you an employee that never sleeps. Here's the honest shape of it.
It is not an unattended bot
It buys you minutes at 2am and clears the questions you've answered a hundred times. It does not run your sales. Anything it can't answer from your knowledge base should reach you — that's the point of the follow-up list, not a fallback we're apologising for.
It's only as good as your text box
An empty knowledge base makes a chatbot that says nothing useful. It won't invent your prices to fill the silence — which is the behaviour you want, but it means the setup is the work. Twenty minutes of writing beats any amount of AI.
Automating WhatsApp Web carries risk
It isn't something WhatsApp officially supports, so there's real terms-of-service risk and no tool — ours included — removes it. Replying to people who messaged you is the low-risk end of this; blasting strangers is not.
AI is off until you turn it on. When it's on, a reply sends only the chat it's answering in — capped at the most recent 40 messages and 24,000 characters per request, and nothing from your other conversations. We never store your chat contents.
And the ones it can't answer?
The hard questions come to you
A chatbot that answers everything is a chatbot that invents things. The value isn't in answering 100% — it's in what happens to the 20% it shouldn't touch.
Needs you 3
AKAhmed K.4d silent
MRMaria R.2h waiting
CWChen W.promised
They land on a list, not in a void
The negotiation, the complaint, the customer who went quiet after your quote — those are yours. Wanova tracks who's still waiting and what you promised, and drafts the reply when you get to it.
The list runs locally — no AI requiredYour credits, in plain sight
The account panel shows exactly what you've used this period and what the engine used on your behalf — two separate meters, so automation never quietly eats the replies you paid for.
It runs on your plan's monthly credits: 15,000 on Professional and 35,000 on Expert, shared across every WhatsApp number on the account. One answer costs one credit. Starter doesn't include AI answers.
No surprise overage
Plans
The chatbot starts on Professional — the credit pool grows
One price per account — not per person and not per number. Every plan covers several WhatsApp numbers and as many teammates as share them. The AI co-pilot starts on Professional.
Starter$14.99/moAI answers start on Professional
Professional$24.99/mo15,000 AI credits/mo
Expert$34.99/mo35,000 AI credits/mo
AI is opt-in and discloses what it sends · See full pricing
FAQ
Questions, answered
Does the chatbot reply on its own, or do I approve it?
Both, and you choose per rule. In draft mode it writes into your composer and nothing goes out until you press send. In auto mode it can answer by itself — but every reply is held for a few seconds in plain sight first, and if you start typing in that chat it's cancelled and you take over.
Where do the answers come from?
From a knowledge base you write yourself — a plain text box where you put your prices, policies, shipping rules and the questions you answer constantly. Numbers and promises in a reply have to trace back to that text, your own notes, or something actually said in the chat. It's not a general-purpose AI that invents an answer when it doesn't know.
Can I test it before it talks to a real customer?
Yes. The agent editor has a test chat: you play the customer over as many turns as you like and watch how it would reply. It doesn't touch your real conversations, doesn't save anything, and sends nothing.
Is this an unattended bot I can leave running?
No, and we'd rather you didn't treat it as one. It buys you minutes at 2am and clears the questions you've answered a hundred times. Anything it can't answer from your knowledge base should reach you rather than be improvised — and the follow-up list is where those land.
What does the chatbot send to the AI provider?
The knowledge base you wrote, the chat it's answering in, and — if you keep customer records — your notes on that contact. The chat is capped at the most recent 40 messages and 24,000 characters per request, and nothing from your other conversations goes with it. AI is off until you switch it on, the extension discloses what was sent and to which provider, and we never store your chat contents.
Which plans include the chatbot?
Professional and Expert. Starter is the local messaging toolbox — broadcasts, scheduling, templates, keyword replies, exports and the privacy tools — plus a taste of AI translation and voice. Understanding what someone meant costs us model time on every message, so it starts on Professional.
Related
Keep reading
Answer the easy ones. Keep the ones that matter.
Try every feature free for 7 days. No credit card.
Start free trial