Note /
An agent is not a chatbot
The difference is not how well the thing writes. It is whether it can act, and almost everything else about this service follows from that.
A chatbot answers, an agent acts
A chatbot takes a question and returns text in its own window. Whatever it produces, a person carries across into the tool where the work actually happens. An agent is given access to those tools directly, so what comes back is the work rather than a description of the work: a record written into your CRM, or a report assembled on a schedule while nobody is watching. The connector half of the add-on menu is the plainest description of this difference we can give you, because each of those items is another system an agent is being given a way into.
It lives where your team already works
An installed deployment is not one more tab to remember. A Team Install puts a multiplayer-class deployment on a Slack or Teams surface with scoped workspaces, so the work happens in the channel people are already sitting in. A Solo Install is a single personal-class agent fitted to one person’s workflow. The panel on the home page is a scripted illustration of what that looks like in a channel, written rather than recorded, because we have no client transcripts to show and will not invent any.
Acting is what makes hardening necessary
A chatbot that gets something wrong has said something wrong. An agent that gets something wrong has done something wrong, in a system of record, possibly to a customer. That is why the deployment ships with a hardened posture as the default rather than an upgrade. Least privilege comes first, so an agent reaches the tools its job needs and nothing else, and actions with consequences wait for human sign-off. An agent can draft the email or stage the change, and a person approves it before it leaves. What your deployment is allowed to do on its own is written down in the signed handover document, not left to be discovered.
Acting is also what makes maintenance necessary
Something that only answers can be left alone. Something wired into your accounting system and your chat platform is exposed to both of them changing underneath it, and to the framework itself changing too. Clients are pinned to stable versions, and updates are staging-tested before they reach any customer. That is the work a retainer pays for, and the reason this service does not end at installation.
Where the line honestly sits
Not every job needs an agent. A question that only ever needs an answer is served perfectly well by a chatbot, and there is no sense installing anything for it. The case for an agent begins when the answer has to become an action inside a tool you already run, often enough to matter, with someone accountable for it going right. That is why the enquiry form asks what you want agents doing before anything gets recommended, and why answering “unsure” on the form is treated as useful information rather than a gap to fill in with the larger product.