LucyIQ is built on a simple principle: your AI agent only knows what you give it. It answers only from your own uploaded documents and your own website content — Professional and above can optionally search the web, and say when they have. It doesn’t invent facts about your business, and it doesn’t pull answers from other customers’ data.
This is what “private RAG” (retrieval-augmented generation) means in practice:
Each agent has its own private partition of our knowledge database, sealed off at the database level. Your documents are indexed only for your agents and are never shared across accounts.
Every client runs in their own separate workspace. One customer’s content is never visible to another.
Because the agent retrieves from your content before responding, it stays on-topic and avoids the “confident hallucinations” that make general-purpose chatbots risky for business use.
The documents you upload, the website content you choose to index, your agent configuration, and conversation logs so you can review and improve your agents.
To generate a reply, the relevant snippets of your content are sent to our AI processing layer and to our enterprise-grade AI language model provider (named in the full sub-processor list we provide on request), purely to compose that single answer (and to generate the search embeddings that power retrieval). Content sent through the provider’s API is not used to train its models, under that provider’s API terms — this is distinct from consumer chatbot apps, which LucyIQ does not use.
We are located in New Zealand. Your application and its self-hosted automation layer run on dedicated virtual servers in Asia-Pacific, operated by Hostinger. Your knowledge base’s vector database is hosted in Sydney, Australia (AWS ap-southeast-2) via Supabase. All traffic is served over encrypted HTTPS (TLS, HTTP/2 and HTTP/3).
When you add a document, it goes straight from your dashboard into your own private knowledge base — indexed within the same infrastructure that hosts it. We recently removed an external cloud-drive step that files used to pass through, so today there are fewer systems in the pipeline and one less third party that ever touches your content. Indexing is event-driven, so new material is usually searchable within seconds.
Call Assist is an optional add-on that shows your staff answers from your own documents while they are on a phone call. Nothing a caller says is turned into text until your staff member has told them an assistant is listening and confirmed in the call window that the caller agreed. The sound of a call is never stored. It is converted to text by our speech-to-text provider (under API terms that do not allow training on it) and discarded. No transcript of a call is kept, and there is no setting that keeps one: the text is deleted once the summary of the call has been written, and no later than when the call is filed. Our AI provider may hold the text it processes for up to 30 days before deleting it; it does not keep call audio, and it does not use this text to train its models. We do not identify anyone by their voice. Call Assist is not offered in the United States.
You can delete uploaded documents at any time from your dashboard. If a subscription lapses, the account enters a 30-day grace period (site stays accessible, reminder emails sent), then a 30-day locked period (login disabled, data retained) — data is permanently deleted 60 days after your last paid period ends if the subscription isn’t reactivated.
We want your chat agent to work for as many of your visitors as possible — including people who use a keyboard, a screen reader, or other assistive technology. We build and test our embeddable chat widget to follow the internationally recognised Web Content Accessibility Guidelines (WCAG 2.1, Level AA), and we treat accessibility as an ongoing commitment, not a one-off checkbox.
The Franchise plan runs many locations under one franchisor account, with a data model built so a network can be run centrally without surveillance:
We take vulnerability reports seriously. If you believe you’ve found a security issue in LucyIQ, email security@lucyiq.com with details — we’ll acknowledge your report and investigate promptly.
Security or privacy question we haven’t answered? Email security@lucyiq.com and a human will get back to you.
Give us your website and we’ll build you a working AI agent trained on it — in about a minute.
Free · about 60 seconds · no signup