Chat
It answers from your content, or it says it cannot.
Chat is a widget grounded in your own site and documents. It indexes what you point it at, answers questions using only that material, and puts a citation next to each claim so the visitor can check it. Asked something your content does not cover, it says so — because a confident wrong answer to a prospect is worse than no answer.
Priced by conversations a month
- 1,000 conversations / mo€190
- 5,000 conversations / mo€490
- 15,000 conversations / mo€1,200
How it works.
The mechanism, not the marketing. If any of this turns out not to match what the product does, the page is wrong and we want to know.
Point it at your content
Give it your site and it crawls the pages you allow, respecting robots and staying on your own origin. Upload PDFs — pricing sheets, security documentation, specifications — and it reads those too.
Content is chunked and embedded
Pages are split into passages, embedded, and stored with a content hash so re-indexing only touches what actually changed.
A question retrieves passages
The question is embedded and matched against your corpus. The best passages, deduplicated so one page cannot dominate, are assembled into the prompt.
The answer cites its passages
Retrieved content is fenced as untrusted input so a page cannot inject instructions. The answer carries inline markers pointing back at the source pages.
It can act, not only answer
Mid-conversation it can offer real availability, book the call, take a callback request, or remember a fact about this visitor for next time — each of which you can switch off.
What you get.
Refusal as a feature
Out-of-corpus questions get an honest refusal and an offer to put the visitor in touch with a person. It does not improvise.
Citations the visitor can open
Every claim links to the page it came from. Visitors check them, and so should you when you are reviewing transcripts.
Attached to the visitor record
The conversation sits in the same timeline as the pages this person read before it, so a rep gets context rather than an isolated transcript.
Origin-locked
The widget key only works from domains you register. Per-instance rate limits stop a single origin from running up your meter.
Your prompt, your tools
Override the system prompt for tone and policy, and disable any tool you do not want it to reach for.
Transcripts and export
Every conversation is stored, searchable and exportable, with the citations it gave.
<script async src="https://antiqore.com/embed/chat.js"
data-chat-key="ck_live_…" data-title="Ask us"></script>What Chat does not do.
Every product page on this site carries this section. It is the fastest way to be believed about the rest.
- It knows everything about your business
- It knows what you indexed. If a policy only exists in someone's head or a private document you did not upload, it does not know it and will say so.
- Unlimited conversations
- Each conversation costs real model time, so it is metered. Reaching your limit falls back to a capture form rather than silently billing you.
- It replaces support
- It answers what your content covers and hands off what it does not. It is a first line, not a support team.
Questions.
What counts as a conversation?
One visitor session with the widget, however many messages it contains. Six follow-ups is one conversation, not six.
How fast does it pick up content changes?
Re-index on demand or on a schedule. Content hashing means unchanged pages are skipped, so a re-index of a large site is cheap.
Can it speak other languages?
It answers in the language it is asked, from whatever content it has. If your corpus is English-only, its answers will be too.
Chat writes to the same record as the other five.
That is the only reason buying more than one is worth anything.