What Scriptus does — and what it does not pretend to do

You send source material. We handle ingest, chunking, prompts, embed, branding and upkeep. Everything in the admin is set up before you open it, and we change it when you ask.

Trade-off There is no self-serve path today. You cannot sign up and build one yourself.

.txt, .md, .markdown, .html, .htm, .pdf, .docx and .pptx, up to 50 MB a file. Text is extracted and split into passages.

Trade-off Legacy .doc, .ppt and .rtf are rejected. A PDF with no readable text — a scan — is rejected with a message rather than ingested empty. Send those to us and we handle them by hand.

Group source material into collections, assign each to the audiences it belongs to, and grant it to trial visitors, to members, or to neither.

Reader groups can be granted different collections from the same corpus, each with its own prompt. Retrieval searches only what that audience has been granted — the separation is in what gets searched, not only in what the bot is told.

A panel beside the answer shows the work title, the chapter and a short passage — up to three sources per answer, with a page number where the source has one. Citations can be hidden per tenant.

Trade-off Page numbers come from paginated sources. Podcasts, web pages and slide decks often carry none.

When your library has nothing to offer on a question, the bot says so instead of answering from general knowledge. Asked about something outside its texts, Ask Nietzsche answers: "The material I have does not adequately address this question. I will not answer from general knowledge. I do not have related material to offer for this question." Where retrieval comes back empty, the refusal happens before the model is called at all.

Trade-off Refusal is governed by your prompt as well as by retrieval, so it is a strong default rather than a hard guarantee on every question. We are tightening it.

Readers keep named threads, titled from their first question, and can return to them. History can be turned off per tenant.

Answers can be read aloud on request, using your own ElevenLabs account and chosen voice. Audio is generated when a reader asks for it, then cached.

Trade-off Off by default, and it needs your own ElevenLabs account. Text is the primary surface.

Opening tiles and rotating greetings you write, assigned per audience, plus a first-visit note in your own words.

Email magic link, or single sign-on from WordPress or Kajabi.

Coverage across your material, the shapes of question people ask, topics drawn from the chapters actually retrieved, which documents were used, which audience asked, and answer quality.

Trade-off These reports carry classifications only — no user IDs, no email addresses, and never a reader's question as written. Where no source label matches, a topic may be a few de-identified words taken from the question. Counts below 25 records are shown with a note that they are not a basis for action.

What readers asked that your library could not support — the subject, and whether nothing matched or the chapter was too thin. Read beside which sources were retrieved, heavy use against a thin chapter tells you what to write next.

Trade-off A gap is a template sentence built from the question's shape and topic. It never carries the reader's own wording.

The same report over 7, 30, 90 days or a custom window. Download as Markdown or print to PDF.

Verbatim conversations from your workspace, inside a rolling window set between 1 and 90 days. Off until Scriptus switches it on for your workspace — your own owners and admins cannot enable it or change the window themselves. Every thread opened is recorded.

Trade-off Older conversations stay in the database but cannot be opened here. The access record is written but there is no screen for reading it yet, so we cannot hand you an access report today.

You connect your own language model and ElevenLabs accounts on the Connections screen, and those providers bill you directly at their rates. Scriptus does not resell or mark up usage. Stored keys are encrypted at rest.

Trade-off Until you connect your own accounts, a workspace we build for you may run on ours. Your usage also varies with how much your readers use the bot — the $500 is the platform fee, not the total cost of running it.

Set a free question limit per visitor, or leave it unlimited. When the limit is reached the reader sees your subscribe options.

Trade-off The count follows a visitor's session. Someone who clears their cookies starts again.

Readers pay on your own Stripe, WooCommerce, Kajabi or Whop account. Card details never touch Scriptus.

Scriptus takes no percentage of what your readers pay you.

Put your offers in the chat, track clicks, and confirm conversions by signed webhook, CSV import or a commerce grant.

Trade-off A click is reported as assisted, never as a closed sale. The two are counted from different tables and are never merged on any screen.

Embed on your own site with a snippet, or share a link. Inline or as a floating launcher with your label and logo.

Your logo, your colours, your launcher label. Visitor-facing copy never mentions Scriptus.

You choose the language model your workspace runs on and connect your own account. The same applies to the voice provider.

Owner and admin run the workspace. Narrower roles cover someone who should only manage knowledge, or only read reports. Support and end-user roles cannot enter the admin area at all.

Your workspace cannot see or reach any other. Scoping is enforced on the server rather than in the browser: a request naming another workspace is refused or scoped back to your own. A Scriptus administrator working inside your workspace is held to that same boundary.

Trade-off Scriptus platform access is one named individual, not a support team. Working from our own workspace he can administer across tenants — that is how support, provisioning and transcript review are switched on.

See who has access, scoped to your workspace only.