The same agent graph
Routing, extraction, conversation and tool execution are the same nodes in every product. What changes is the set of tools an agent is allowed to call.
Products
Status here means what it says. Production is people outside the company on real data. Internal is production on our own operations only — we are the tenant. Development is being built: the interface stands and the engine behind it runs, but nobody outside uses it yet. Research is a line we work on without promising a date — listing it as a product would be the easy lie, so we do not.
Products on infrastructure we operate. Two of them serve users today; the third is in development and its card says so.
An AI layer over enterprise resource planning. A business talks to its own system by text or voice, and the agent carries out the operation under a deterministic control layer that records intent before acting.
A platform for launching and running sites, not a page builder. Multi-page templates with a theme system, editing in place, publishing to our own CDN — and the parts builders leave out: an SEO engine, analytics, file storage and leads in the same admin. An existing site can be moved in rather than rebuilt.
An agent bound to one person rather than to a company. Memory that accumulates over years instead of resetting each session, reachable by voice, and able to act on that person's behalf under the same control layer the enterprise products answer to.
Three different markets on one engine. That is the point of building the engine first — a new product is a configuration, not a rewrite. It inherits the graph, the control layer, memory, the voice channel, billing and the infrastructure; what it adds is its own tools, its own knowledge base and its own interface. That is why the work left on any one of them is specific enough to list rather than an open question.
Routing, extraction, conversation and tool execution are the same nodes in every product. What changes is the set of tools an agent is allowed to call.
Whether the action is issuing an invoice or publishing a site, it passes the same gate and leaves the same kind of record. One place to review, one place to change the rules.
Customer data, memory and retrieved knowledge are separated per organization by design, not by a filter applied afterwards. Nothing crosses between customers.
Lines we work on that are not products. They are listed because they explain where the engine is going, not because they are for sale.
What people ask before deciding which of these is relevant to them.