An agency managing multiple client sites with Chatterbox keeps every client's chatbot answers separate. Each client's knowledge base stays isolated from every other client's, and agency controls let one person manage all the sites together.
You look after five, ten or thirty client websites, and every one of them is a candidate for a chat widget. Before you put it on the whole book, you want a straight answer to one question: what happens when Client A's pricing and Client B's pricing describe two completely different businesses, inside what is, for you, the same tool?
Does one client ever see another client's answers?
No. Each client's knowledge base is isolated from every other client's. Whatever a visitor asks on Client A's site is answered only from what Client A has published or supplied — nothing crawled from Client B's site, nothing uploaded to Client B's dashboard, nothing a Client B visitor has ever asked, can surface in a Client A conversation. That separation is per client, not per plan or per person, so it holds whether you run two client sites or thirty.
What agency controls let you do
Agency controls let one person manage several client sites together, rather than treating each client's chatbot as a disconnected setup you re-learn every time you touch it. That "one person" is you: the same login carries across the accounts you administer, so adding a chatbot to a new client's site is not a separate onboarding you invent from scratch each time. The isolation above and this convenience are not in tension — the sites stay administratively together for you while their content and conversations stay apart from each other.
Installing across sites you don't all build the same way
Your client book is rarely one platform. Some sites are static HTML from years ago, some are WordPress on shared hosting, some were built with a proper deployment pipeline. Chatterbox installs the same way regardless: one script tag pasted before the closing </body> tag, no plugin, no npm install and no build step, and the widget stays under 15KB. That matters more at agency scale than on a single site — you are not maintaining thirty different installation procedures, one per client's stack.
What happens on a client site whose content isn't ready yet
Rolling a chatbot out across a whole book rarely means every client's source material is equally complete on day one. Chatterbox builds its knowledge by crawling each site's own pages, FAQs and documents, so a client whose site is thin on published detail gets a chatbot that reflects that: it says it cannot find the answer rather than inventing one, and it asks the visitor for contact details instead. That behavior is per client too. A newly onboarded client with gaps in their content does not degrade the answers on a client you finished preparing months ago — each site's chatbot is only ever as complete as that site's own published material, and only that site's.
Deciding how you roll it out
Because the data stays separate per client and the management stays together for you, the real rollout question is not technical — it is which client's content is actually ready to be answered from. A client whose FAQ and pricing pages are current is ready today. A client whose pricing lives in a PDF nobody linked, or whose service pages contradict each other, will get a chatbot that mostly asks for contact details until that gets fixed. Treat the rollout as client-by-client content readiness, not as one all-or-nothing switch, and every site you add keeps answering visitors outside business hours without needing you to staff it.