One principle, and everything follows from it
No source, no answer.
Why we built it
Because most tools on the market make the same mistake: they answer at all costs. A language model cannot say no by default — with no information, it invents something that sounds plausible. In customer support that is not a small inconvenience but a business risk: the wrong price, the wrong deadline, the wrong warranty term, and you are the one explaining it afterwards.
What we turned that into, in code
The assistant does not answer from its "world knowledge" but from what you approved. If a question is not covered, it says so. Every answer carries its source as a clickable link. The knowledge base is not closed by the machine but by you: where two of your pages contradict each other, you decide — and what you write is not overwritten by a later re-crawl.
How we work
Every claim that reaches this page is measured first. The numbers you see come from our own test set, and we say that they are ours — they are not customer references. You will not find invented testimonials or logos here either: a product built against hallucination cannot open its introduction with a lie.
Where we are
The product works and is measured, but it is young. If you join now, you will have a direct line to the person building it — and your feedback genuinely shapes what gets built next.
What we believe
"I do not know" is an answer too
In fact it is the most valuable one, when the alternative is guessing.
The knowledge is yours
You approve it, you rewrite it, and you can take it with you any time.
No claim without measurement
What we cannot show in numbers, we do not write down.
See it on your own site
Give us the address, run the crawl, and see what it could answer from your content. If it does not convince you, you have paid nothing.