Composable commerce & headless storefronts

AI Chatbot for commercetools: one script in the storefront your team already owns

Q: How do we install a chatbot on a commercetools storefront?

The storefront is your own application, so the widget is installed the way any front-end script is installed: your team adds the Asyntai snippet to the shell that renders every page, commits it, and ships it through the normal pipeline. There is nothing to configure on the commercetools side, because the widget never touches the API layer.

Q: Is there a commercetools app or plugin for the chatbot?

No, and there should not be. commercetools is headless: it serves commerce APIs and does not render your HTML. A plugin would have nothing to plug into. Installation belongs in the storefront repository your team controls, which also means you decide exactly which routes and locales load the widget.

Q: Where does the snippet go in a commercetools Frontend project?

commercetools Frontend delivers your site with Next.js from your own repository, so the snippet goes in the root document or layout that wraps every route. Teams that prefer business users to control placement can instead wrap it in a small custom component and register it, so it can be added through Studio.

Q: Does the chatbot work with a fully custom storefront on the commercetools APIs?

Yes. The widget is a framework-agnostic browser script with no SDK and no build step, so it runs in Next.js, Remix, Nuxt, Astro, Angular or a hand-rolled front end. If your organisation already deploys tags through a tag manager, that route works as well.

Q: How does the chatbot get product, pricing and policy information?

Asyntai crawls your published storefront and any other URLs you point it at, such as a help centre or a specification library, and you can upload documents and write custom instructions. The bot answers using your own content instead of general AI knowledge.

Headless means nobody hands you a plugin. Your front end is your code, your pipeline, your review process — and that is exactly why adding conversational support here is a small job rather than a project. Asyntai is a single browser script with no SDK and no coupling to the commerce APIs. Your storefront team drops it into the shell that renders every route, and the site starts answering buyers from your own product, pricing and policy content.

Get Started Free See the install path

See the chatbot answer from your commercetools storefront

Enter a storefront URL — production or a public staging environment — and see how the bot handles your catalogue and policy pages.

Why a composable stack makes the chatbot question awkward

Enterprises move to commercetools to stop being told what their storefront can look like. Catalogue, cart, pricing, promotions and orders become APIs; the experience layer becomes a normal application maintained by a normal engineering team. That trade is deliberate and it works.

It also changes what "adding a chat widget" means. On a monolithic platform, somebody installs an extension from a marketplace and the vendor's code lands inside the storefront runtime. In a commercetools programme there is no such surface. The storefront team owns the HTML, the render pipeline, the content security policy and the performance budget. Any third-party script is a deliberate decision, reviewed like any other dependency.

Owning the front end is the whole point. It also means nobody else can install anything into it.

Meanwhile the support problem in these programmes is unusually large, because the knowledge is unusually scattered. Product data sits in commercetools or a PIM. Freight rules and lead times sit in an ERP. Contract terms sit in a PDF that account managers email. Returns policy sits in a CMS. A B2B buyer comparing two part numbers, or asking whether an item ships to a specific country on a net-30 account, is asking one question that spans four systems — and typically gets a contact form and a two-day wait.

A retrieval-based chatbot is a practical answer because it does not need those systems merged. Point it at the published storefront and the help centre, upload the specification sheets and the terms document, write instructions for what it must never guess at, and it answers from that material. Anything outside it becomes a captured lead with the full transcript attached, routed to the person who can actually respond. B2C shoppers get the delivery and returns answers that otherwise become tickets; B2B buyers get a first response at midnight instead of Tuesday morning.

How the chatbot is deployed on a commercetools storefront

Create the bot and point it at your content. Sign up at Asyntai, enter the storefront domain so the crawler can read your live pages, add any extra sources — help centre, spec sheets, terms — and copy the snippet from the dashboard.

Add the snippet to the storefront shell. With commercetools Frontend, delivery runs on Next.js out of your own repository, so it belongs in the root document or layout that wraps every route. On a bespoke front end built directly against the APIs, it belongs in the equivalent base template. One place, every page.

Ship it through your normal pipeline. Commit, review, deploy to staging, verify against your content security policy and performance budget, then promote. Nothing is configured inside commercetools itself, because the widget never talks to the commerce APIs.

If business users should control placement. Wrap the snippet in a small custom component and register it, so it can be placed and configured from Studio like any other component. Teams that already run an enterprise tag manager often deploy it there instead, which keeps the change out of the release train entirely.

Two consequences of the headless model are worth stating plainly. First, there is no commercetools marketplace app to install and no vendor code inside your commerce layer. Second, because you own the injection point, you decide which locales, which storefronts and which routes load the widget — a B2B portal and a B2C shop running from the same project can behave completely differently.

What the chatbot handles on a commercetools storefront

Answers using your own content

Crawled storefront pages plus uploaded specification sheets, freight terms and policy documents. Retrieval over your material, so it does not improvise a lead time.

Qualified handoff for B2B enquiries

Quote requests, account questions and contract terms become captured leads with the full conversation attached, ready for the right team to pick up.

30+ languages for multi-market stores

One deployment serves every locale you sell into. The bot detects the visitor's language and replies in it, with no per-market configuration.

No coupling to your architecture

A plain browser script with no SDK and no build step. Replace a component library or re-platform the front end and the widget keeps working.

commercetools chatbot questions

How do we install a chatbot on a commercetools storefront?

Your team adds the snippet to the shell that renders every page and ships it through the normal pipeline. Nothing is configured on the commerce side.

Is there a commercetools app or plugin for the chatbot?

No. Headless means commercetools serves APIs and does not render your HTML, so there is nothing for a plugin to attach to. Installation lives in your storefront repository.

Where does the snippet go in a commercetools Frontend project?

In the root document or layout of the Next.js delivery app in your repository. Alternatively, wrap it in a registered custom component so it can be placed from Studio.

Does the chatbot work with a fully custom storefront on the commercetools APIs?

Yes. It is framework-agnostic — Next.js, Remix, Nuxt, Astro, Angular or plain HTML. A tag manager deployment also works.

How does the chatbot get product, pricing and policy information?

It crawls the published storefront and any URLs you add, and you can upload documents and write custom instructions. Plan limits are on the pricing page; the free plan includes 100 messages a month for a proof of concept.

Other systems in a composable stack

The same script drops into any front end your team controls.

Prove it on staging before it reaches the release train

One script tag, one layout file, no change to your commerce layer. Start free with 100 messages a month and see how the bot answers from your real catalogue content.

Get Started Free