What you see here is the demo: this website has no chatbot widget — it is the agent. The chat on the homepage answers from Creaminds' curated knowledge base, citing the source for every answer.

What "Agent-First" actually means

The chat is the main navigation — no "?" bubble in the corner, no separate bot. The hero section is the input field.
RAG over a curated knowledge base, not the open web. If something isn't covered, the agent says so explicitly.
Local model, on-premises inference. No request leaves the datacenter, no external LLM provider sees the prompts.
Stateless. No user tracking, no stored history, no chat cookie.

How it works

Creaminds' approved Wiki articles are converted into a vector index. For each query, the agent picks the relevant sources, formulates the answer, and cites the source. Model, RAG stack, and vector DB all run on our own hardware. Details: Technology · LLM Self-Hosting.

Who this is for

Mid-market companies that want to make their documentation, FAQ, product knowledge, or internal wiki accessible as a chat interface — for customers, sales partners, or employees — without data leaking to external providers.

Demo

The homepage is the live demo of this concept.