Chat with your website
Enter a URL. We crawl a few pages, build a retrieval index on the fly, and you ask the site anything. This is retrieval-augmented generation (RAG), the pattern behind serious AI assistants, running on your content in seconds.
The index lives in memory for ~20 minutes, then it is gone. Rate-limited; nothing stored permanently.
What happens under the hood
Crawl
The homepage plus up to five of the site’s own most informative pages.
Index
The text is split into chunks and indexed for keyword retrieval (BM25), live, in seconds.
Chat
Questions are answered only from the indexed pages, with the sources shown.
Why it matters
This is the same RAG pattern we build into support, intranet, and product systems.
This demo, on your real data.
We build the production version: grounded chat over your documentation, product data, and CRM, inside the systems you already use.
From the blog
- Prevent AI hallucinations: answers need evidence
How to prevent AI hallucinations: why agents should answer only with evidence from your data, how that is built, and what happens with contradictions.
- What is an LLM wiki?
An LLM wiki is a knowledge base of linked Markdown pages that a language model writes from raw sources. Karpathy's pattern, and how it compares to RAG.
- What is RAG? Retrieval in plain terms
Retrieval-augmented generation (RAG) lets an AI answer from your own documents: find the right passages, then write a grounded answer that cites them.
