HiQBot is an AI chatbot for business websites, added with one snippet. It answers from documents you upload, works out whether the visitor is buying, and books the meeting instead of promising someone will be in touch. It does that at three in the morning and on a bank holiday, and it does it on the first enquiry and the four hundredth without getting slower. Plans start at $49 a month and the widget is on all of them.
Installing it is one tag in your <head>. Most teams have it answering real questions the same day they decide to try it.
Most site chat is a form with a typing indicator. This one reads the documents you gave it and answers from them, which is the whole difference. When it does not know, it says so rather than guessing, because a confident wrong answer costs you more than no answer.
What it knows is exactly what you upload. No crawler reads your site in the background and nothing syncs from a wiki, so nothing goes stale without you knowing which file went stale. When your refund policy changes you replace the file, and the next visitor gets the new answer.
It works the same on a marketing site, a documentation site and a signed-in app. On the app it can confirm the visitor is the account your system says they are, which is the difference between answering a billing question and refusing to.
Paste a single tag into your site and the widget is there. No package to install, no framework to fight with, nothing your developer has to put in next sprint.
Replies stream word by word instead of landing after a spinner. Visitors stay in the conversation rather than deciding it’s broken and closing the tab.
The widget follows your site’s light or dark theme, so it reads as part of the page instead of a third-party box someone bolted on.
Upload your help articles, spec sheets, and policies. The agent answers from those and says when it doesn’t know, which is the difference between useful and embarrassing.
Signed identity verification lets the widget confirm a logged-in visitor is who your app says they are, so support conversations don’t start with twenty questions.
From the Starter plan up, the agent brings in a rep who is on shift and passes over the full conversation with it.
What it costs
We publish AI chatbot pricing. Nothing here is quote-on-request. What you pay more for higher up is seats for the team, the extra channels and a bigger allowance, never the agent itself.
$49 /month
2,500 credits a month
The agent runs your website chat, WhatsApp and Instagram on its own. One seat, no live-chat seats, so nobody has to sit in the inbox for it to work.
$139 /month
6,000 credits a month
Adds two live-chat seats, so the agent can hand a conversation to a person who is on shift. Messenger, Telegram and LINE turn on here too.
$229 /month
10,000 credits a month
Five seats, and the support agent with tickets behind it. This is the tier for a team that answers as well as sells.
A plan sells message credits, and one AI reply draws credits according to the depth you set for that inbox. A Core reply uses one credit. Balanced uses four and Max uses eight, because those levels think for longer before they answer. You choose the level per inbox, so the inbox that closes deals can run deep while the one answering opening hours runs cheap.
Credits are paid up front and nothing runs over. Replies pause at zero until the plan renews or you top up, so no bill arrives that you did not expect. On monthly plans, whatever you do not spend carries into the next period and stays there for two billing periods. See the full pricing breakdown, or work out what a message actually costs on WhatsApp if that channel matters to you.
Prices re-read from our plans API on 9 September 2026. This page was published on 25 July 2026 and last updated on 9 September 2026.
Tuned to your industry
The agent ships with five industry packs. Each one has its own qualifying questions and its own closing move, written for how that industry’s buyers actually talk. A guest asking about a dietary requirement and a buyer asking about finance want different things, and the pack decides which questions come next.
The pack is not a script. It changes what the agent treats as a buying signal, which questions it asks before it commits to anything, and what counts as a finished conversation. In a restaurant a finished conversation is a table. In a dealership it is a test drive. The same visitor message means different things in those two businesses, and the pack is what tells them apart.
You pick your industry when you set the agent up, and you can hand it your own wording for anything it gets close but not right.
The agent asks what they drive now, what they are replacing it with and when they want it, then puts a test drive in the diary. A finance question gets a straight answer, not a form.
Dates, party size, what the room includes, what the kitchen can do about an allergy. It knows your opening hours and it takes the reservation rather than telling the guest to call.
Menu questions, dietary questions and a table booked at the end of them. The same pack handles an order the way it handles a reservation, because to the guest those are one conversation.
It works out where they want to go, when, how many are travelling and what they expect to be included, then books the call. Your consultant opens the conversation already knowing the shape of the trip.
It answers on plans, limits and what happens when someone upgrades. When a visitor is clearly evaluating rather than reading, it books the demo and stops explaining.
All five packs run on the Solo plan, so the industry-tuned agent starts at $49 and not a tier higher. Read the five industries the agent is tuned for in full, or ask us to extend a pack for how your business runs.
The closing move
An AI chatbot that answers well and ends there has moved the conversation nowhere. Every released industry pack closes on a booking, which is why AI chatbot appointment booking is the default outcome here rather than a per-vertical extra. The agent sets the appointment itself and does not hand the visitor a form and hope.
Someone who books and then cannot make it can cancel or move it in the same chat. That is not gated to a higher plan, because a cancellation is as often a sales conversation as a support one.
Each industry pack has its own qualifying questions, written for how that industry’s buyers actually talk. A visitor who is browsing gets an answer. A visitor who is buying gets asked the things your team would ask.
Date, time, name, whatever else that booking needs. The agent validates what it collected before it commits anything, so you are not confirming a meeting against a typo.
It writes against the contact, next to the conversation that produced it. The team sees what was said before they see what was booked.
Connect Google Calendar and every meeting the agent takes also lands on the team’s real calendar. Available from Solo.
Bookings are written inside HiQBot first, against the contact and next to the conversation. A connected Google Calendar mirrors them onto your team’s calendar.
When a person should take over
The agent spots that it is out of its depth, checks who is on shift and hands the whole conversation over with everything already said attached. Your rep opens it knowing what the visitor asked and what the agent already told them, so nobody starts again from the top.
This needs a live-chat seat, so it is Starter and above. On Solo the agent runs the whole conversation on its own, which is the point of that plan.
You set what triggers it. Some teams hand over on any pricing question above a threshold, others only when the visitor asks for a person. The agent also checks whether anyone is actually on shift before it promises one, so nobody is told a human is coming at 2am.
Some conversations you want a person on from the first message. Set that channel to live chat and the AI stays out of it entirely. Credits are drawn by AI replies, so a channel the team answers does not draw them at all.
Switch intake on and the agent collects the details first, hands over, and draws a single credit at the moment it reaches a person. Nothing after that. A channel that skips the AI entirely.
Getting it live
Drop the snippet in, the same way you added analytics. It works on whatever you built the site with.
Upload the documents the team already answers from. The agent reads them and starts replying from what you gave it.
Set what the agent handles alone, when it should book a meeting, and what it should hand to a person instead of guessing at.
<script src="https://widget.hiqbot.com" async></script>
<script>
window.WIDGET_CONFIG = {
organizationId: "org_xxxx",
websiteToken: "wt_xxxx",
};
</script>Your dashboard fills in the IDs and gives you the exact snippet to copy.
Website chat, WhatsApp and Instagram run from the first plan. Messenger, Telegram and LINE turn on from Starter. It is the same documents and the same answers behind all of them, in one inbox, so a customer who starts on your site and finishes on WhatsApp is not starting over.
You do not retrain it per channel. Upload once, and the agent answers the same way whether the question arrives on your pricing page or in a WhatsApp thread at the weekend.
Someone lands from a search in Portuguese and gets answered in Portuguese. The team can keep working in theirs, and nobody configures a list of languages or sets anything up per language.
Solo is $49 a month, Starter is $139 and Growth is $229. The website widget and the AI agent behind it are on all three. What changes higher up is seats for your team, the extra channels, and how many credits you get. Prices were read from our plans API on 9 September 2026.
You buy message credits, and a reply draws credits according to the depth you set for that inbox. Solo carries 2,500 credits a month, Starter 6,000, Growth 10,000. A Core reply uses one credit. Balanced uses four and Max uses eight, because those levels think harder before answering. You pick the level per inbox, so a sales conversation can run deep while an FAQ inbox runs cheap.
On monthly plans, yes. Whatever you do not spend from your plan allowance carries into the next period and stays available for two billing periods before it expires. The current month is always spent first and carried credits sit underneath it. Two things clear the bank: annual plans do not carry at all, and a renewal payment that lands after the grace window forfeits everything carried, with no way to restore it afterwards. Credit packs bought as an add-on do not carry.
Someone opens the chat in Portuguese and gets answered in Portuguese. Your team keeps working in whatever language they work in. You do not configure this per language or pick from a list.
You upload files. Help articles, spec sheets, price lists, policies, the PDF your team keeps re-sending. The agent answers from those and tells the visitor when it does not know rather than inventing an answer. No crawler reads your site and nothing syncs automatically from a wiki, so what the agent knows is exactly what you gave it.
The trial runs seven days and takes a card up front, which is not charged until the seven days are up. No free tier sits underneath it, so if you do nothing the plan you picked starts billing. Cancel inside the seven days and it does not.
Put the widget on your pricing page first. That is where the good questions and the quiet exits happen, and every one of those exits was someone who wanted to buy and could not get an answer. The trial runs seven days, takes a card up front and does not charge it until the seven days are up.