AI support for your site — in private preview
Minervia is an AI assistant you embed in your own site. It reads your pages and documents into a curated knowledge base and answers visitors from that knowledge — citing its sources, in the visitor's language. Anything it does not know becomes a ticket for a human.
It either knows, or it escalates.
Every answer carries the sources it was built from. No knowledge, no answer — never a guess dressed up as one.
Unanswerable questions become numbered tickets. Outside business hours, everything does.
Hold every answer as a draft for a human to approve, edit or reject — or only the low-confidence ones.
Minervia crawls your site — sitemap and internal links — and reads the documents you upload: PDF, Word, Markdown. Rescans run on a schedule and reprocess only what changed.
Everything is rewritten into clean, deduplicated entries you review in the dashboard: correct an entry, exclude a page, add a manual answer. What the assistant knows is never a mystery.
A floating chat icon or an embedded component, script-injected with one line. Answers come from the knowledge base and your live data, with citations, in the visitor's language.
Tickets with states — open, pending, resolved. A visitor can leave an email to hear back when a pending answer is approved.
Answers can carry your real numbers — products, prices, stock, an order's state.
Point the assistant at an endpoint of your API — URL, auth, and a description it reads. Timeouts, call limits and a domain allowlist bound every call.
The widget can call your API with the visitor's own session — their cart, their orders. That personal data never touches Minervia's backend.
A small service you install inside your network. Outbound-only — no open ports — with a human-approved allowlist of read-only tools and a local ledger of every execution. Your database answers; your credentials never leave.
Basic mode writes nothing to the visitor's browser. Support mode writes one opaque random id, only at first chat use — never on page load, never across sites.
A control in the widget clears the id and deletes the linked history server-side. Retention is yours to configure.
Run the language model on your own machine through the Connector. Visitor questions and answers never leave your infrastructure.
We run the model on our provider keys. Usage is billed from prepaid credits at published per-model rates, and a monthly spend cap per agent stops overruns — exhausted, the agent says so and degrades to tickets.
Plug in your own provider account. You pay your provider; we charge nothing for AI.
Your machine, through the Connector, dialing out. No AI cost, and no visitor data leaving your network.
Three things cost $5. AI usage is the only variable — and you can make it zero.
An assistant on a site, with its knowledge base, widget and ticket flow.
A live-data service installed in your network.
A person who reviews sessions, approves drafts, edits knowledge.
All monthly. Plus AI usage from prepaid credits at published per-model rates — bring your own key or run a local model and the AI line is zero. No setup fees, no seat minimums, no annual lock-in, and a spend cap per agent so a busy month never becomes a surprise invoice.