The AI Chat Assistant for Software and SaaS

For providers whose users ask recurring questions about setup, features, plans or status. Your documentation goes into the knowledge base as a website, a file, free text or a maintained question-and-answer pair; a website run reads at most 25 pages per pass and obeys robots.txt, and a PDF is read through its existing text layer. Whatever stayed open is recorded as an unanswered question and as a knowledge gap in conversation analytics — and becomes your next page of documentation. Questions about system or billing status are answered by a tool of your own against your system, reading only and filtered down to the declared result fields. More under assistant for software.

From your own sources: documentation, help articles, changelog, price list and connected status systems. It is bound to this content and instructed not to invent answers into the blue. The search combines a keyword comparison with a local feature calculation, whose rankings are merged; every section keeps its heading and origin, which is why an answer can name the source it used. If it finds no supporting section, it usually says so openly and hands over to a human. Such cases do not vanish: they collect as unanswered questions and as knowledge gaps in the evaluation, and the next addition to your knowledge base follows from them.

The appointment tool type takes a demo appointment as a request that you confirm; the assistant enters it straight into your calendar only with a calendar connection set up in the Custom package. A trial access runs through a tool of your own against your system. For that you describe a schema with up to twenty fields; the input is trimmed to it before execution, and if a required field is missing nothing is executed at all. The access key for your system is stored, replaced by a placeholder in the interface and never returned. Both operations write and therefore require confirmation: the visitor reads the proposal in plain text, and only their click triggers it. How approvals and the connection work in detail is described under tool control.

That depends on your questions, which is exactly why it is counted rather than estimated. The evaluation keeps conversations, self-resolved cases, cases handed to your team and questions without an answer as separate metrics; the weekly report repeats them as labelled cards over the preceding seven days. Every recurring standard question about setup, features or plans that the assistant resolves in chat never becomes a ticket. Where it has to pass, knowledge gaps and unanswered questions appear as their own lists, from which the next improvement follows. We promise no target figure; you see your own in the conversation analytics.

Yes, provided you connect the system. A reading tool may run without a query if you explicitly set it to automatic; everything that writes is forced to confirmation when created or changed, no matter what the request says. The call goes to exactly one stored connection, the target is checked beforehand, the deadline is eight seconds and can be set between one and twenty seconds; at most 64 KiB are accepted and redirects are not followed. What comes back are only the result fields you declared. Contract changes stay outside all this: they run through a handover to a person. The details are described under tool control.

The assistant sits in the Pro package: from 2,900 € net one-time setup plus 149 € per month net operation, including hosting in Germany, care and conversation analytics. Pro covers ten seats, five assistants, 50 knowledge sources and 2,000 documents, which is enough for the manual, the change log and help articles in one shared knowledge base. For your own API integrations, writing actions and individual workflows the Custom package is available from 4,900 € net setup, with operation there by scope. Every project begins with a free initial consultation. All packages are listed under pricing. All services can be cancelled monthly; there is no minimum term.

The frame around it is tight and anchored in the code. The assistant calls only tools you have entered and switched to active; there is no open access to your systems. Anything that writes is forced to confirmation when created or changed, no matter what the request says — without confirmation only something that reads and is explicitly set to automatic will run. Every call leaves a log row that is written before execution and knows seven states, from proposed to deadline exceeded. A one-time key made of conversation, tool and sorted input prevents double execution, and a missing required field prevents execution altogether. At most five calls run per answer. More under tool control.

Not in the same instant, but without you doing anything. Website sources carry a refresh rhythm, seven days by default; it can be set from one day up to 365 days, and the value zero switches it off. The run happens at night. If a change should take effect sooner, trigger the read manually under Knowledge and then check the state of the source to see how many sections were produced. Each run writes a new version and switches over only at the end: if it breaks off, the previous state keeps answering, and a run that wrote nothing deletes nothing either. Five versions are kept. More under knowledge base.

Yes, in the Custom package. We chain individual actions into a continuous process — for example read out, calculate, create a document, create a ticket and hand over the lead. This creates a path without media breaks in which each step has its own permission: anything that does more than read is first shown to the visitor as a proposal and confirmed by them. The chain is bounded so it cannot run away: per answer the assistant performs at most five tool calls, and only one when the monthly allowance is tight. Every single call is then recorded in the log, with input, result, duration and the path by which it was approved.