These are not screenshots or slide decks. Each page below is the real engine — it scores real data, parses real documents, and one of them goes out to the live web while you watch. Press the green button on any page and it runs.
Built with plain HTML and JavaScript, no frameworks and no paid APIs, so they load instantly and work offline (except the research agent, which needs the internet by definition).
Scans a market, scores every business on seven buying signals, throws away the time-wasters and writes a first message that quotes that business's own numbers back to them.
Answers customer questions, looks up real order records, and measures its own confidence. When it is not sure, it refuses to guess and raises a ticket for a human instead.
Reads six completely different invoice layouts with no template, pulls out every field, then checks the arithmetic, the GST number and the duplicates. It catches money being lost.
Watches every deal, works out which ones are quietly dying, sorts today's calls by the money actually at risk, and drafts the follow-up message for each one.
Give it a topic. It breaks the question into parts, searches Wikipedia and Hacker News live, reads what it finds, cross-checks facts against each other and writes a brief that admits what it could not confirm.
A hotel receptionist that checks room availability, quotes a rate, and confirms the booking in a normal conversation. This one also exists as a real WhatsApp bot that books rooms on a live number.
The whole rental in one flow: a rider asks on WhatsApp, gets live fleet availability and an exact quote with km limits, verifies their licence, pays the advance, then the counter hands over the keys and settles the deposit on return.
Riders compare the same bike or car across several rental companies and book in one payment. Each company gets bookings in its own partner app, and a business model tab works out what the platform earns.