DocsChat Pro · live demo

Ask the docs. Keep the answers honest.

The chat bubble in the corner is DocsChat Pro, answering from the sample Nimbus documentation. Use it like a visitor would, then look at what the site owner sees below.

What is real here: the search, the injection guard, personal-data redaction, answer cache, citation and grounding checks, analytics, feedback, human handoff and the owner API all run as the actual DocsChat Pro server code, inside your browser. What is not: the language model is replaced by a tiny offline writer that quotes the retrieved passages and cites them, and the support webhook is replaced by the inbox below. With a real model key, answers are written by Claude (or any OpenAI-compatible model). Nothing you type leaves this page.

Things to try

What the site owner sees

Live analytics from your questions: most asked, gaps in the docs, ratings, response time, cache hits. Try the 👎 button on an answer and leave a comment.

Support inbox

Requests sent with “Talk to a human” arrive at your own webhook (Slack, Discord, Teams, your helpdesk). Here they appear in this box.

Add it to your docs

One script tag; the server side is a Vercel function set or a single Node file. Collections, embeddings, analytics and handoff are switched on with environment variables.