Grounded
Every answer is drawn from a corpus you curate. Claims link back to source passages you can open and verify — not on fluency alone.
Product sheet · Quaestra.ai
Quaestra.ai helps organisations answer hard questions from sources they trust — reasoned through their own point of view, with every claim tied to evidence they can check.
01 · Quaestra.ai overview
Quaestra.ai is built for specialists whose edge is a view the consensus does not hold yet. It draws answers only from a body of knowledge you curate, reasons through the framework you choose, and ties each claim to the exact passage it came from. Private deployments are available so your sources stay under your control.
Every answer is drawn from a corpus you curate. Claims link back to source passages you can open and verify — not on fluency alone.
Questions are reasoned through the expert framework you define, so a minority view is heard on its own terms rather than flattened into the mainstream.
Your corpus remains yours: access-controlled, auditable, and never retained or used to train anyone else’s model — including ours.
Swap underlying models as better ones ship. Your sources and frameworks stay put — you are never locked to a single provider.
02 · Platform capabilities
The Quaestra.ai platform console gives operators a single place to host, govern, and commercialise many independent organisation workspaces.
At-a-glance status of the platform instance and the organisations running on it.
Provision, suspend, delete, and manage organisations. Login as admin to support a tenant without sharing their credentials. Each tenant gets its own workspace, library, users, and data boundary.
Attach and approve custom domains (BYO domain) and white-label hostnames so each organisation can present Quaestra.ai under its own brand and URL.
Per-organisation analytics configuration, including optional session insights (Clarity) where enabled for a given host or custom domain.
Invite and manage platform operators, including status controls for who can run the console.
Review an audit trail of sensitive platform actions — provisioning, access changes, commercial updates, and more.
Configure platform-level commercial settings that connect billing across tenant workspaces.
Enable or disable Embed Chat for a tenant from the platform console, alongside Domains and Analytics for that organisation.
Platform operators and organisation users sign in securely — passwordless email codes, with optional Google or Microsoft sign-in where enabled.
03 · Tenant capabilities
Each tenant is a fully branded Quaestra.ai environment. Administrators manage the shared library, experience, people, models, plans, and integrations from the areas below.
Add documents to the shared expert corpus and organise them in the Library — including tags, visibility, and peer-review marking where used. Administrators can also ingest PDFs by email through Inbound Email.
Technical papers are read with their equations kept whole, and numbered formulas such as Eq. (2.19) are indexed so questions about a specific equation can land on that formula and the surrounding explanation. Answers can quote formulas faithfully and cite the equation number and page. Administrators can browse a document’s equation index, open a labelled equation to jump to that page in the PDF (with the equation number highlighted when the page text includes it), and reprocess a document — or a selection — so existing material benefits from later reading improvements. Documents being read report their progress as a live percentage in the Library, so long papers visibly advance. Math Corpus lets administrators also build independent ground truth and compare extraction quality over time without changing Ask.
Connect a cloud drive as a read-only source of PDFs. Choose folders to watch — or select individual files within a folder — sync new and updated files into the shared library, and keep searching and viewing available even when the drive is temporarily offline. Disconnecting removes the Quaestra copies; source drives are never modified.
Organisation administrators run the workspace from the Admin menu: people, models, plans, analytics, integrations, and configuration. Use View as User to preview the end-user experience without losing admin access.
Configure the Ask experience: assistant name (also used to brand and address the organisation on the shared Quaestra.ai Alexa skill), Q&A tab label and welcome copy, chat disclaimer, suggested questions, chat attachments, Chat Infographics, and how the conversation presents sources.
Logo, login background, browser title, assistant name, and related Configure App branding so the tenant looks like your organisation — alongside Bring Your Own Domain from the platform.
Author the expert frameworks that shape how questions are reasoned — including router settings, defaults, hypothesis routes, and routing analytics — so answers follow your organisation’s lens.
Show or hide the My Documents navigation tab for admins and testers, and edit Virtual Review prompts. Reviews use the organisation Default Model from Models.
Publish public pages at friendly URLs, with intro content, navigation, indexing, canonical URL, social image, sitemap, and robots controls for discoverability before sign-in.
Invite people, assign roles, manage personas and models, export membership (CSV), review per-user Q&A history (CSV/PDF), and wipe or delete accounts when required. Enable or disable each person’s daily and weekly Q&A digests.
Send branded daily and weekly email digests of each person’s own questions (opt-in; off by default), with equations shown as formulas in the message. Set defaults for new users, choose send time (UTC) and weekly day, tune intro and sign-off messaging, cap how many conversations appear, and review engagement such as opens and clicks. When someone has been away and has nothing new to summarise, digests can still nudge them back with suggested questions from your Question Bank.
Choose who may use Amazon Alexa for the organisation from Admin → Alexa: OFF, Admin & Tester, or All Users. Eligible people link the shared Quaestra.ai skill from Profile → Integrations with a pairing code (regular users also need Alexa Access on their plan; admins and testers do not when visibility includes them). Assistant Name brands voice replies and routes multi-organisation questions inside the skill. Longer questions may ask you to say continue for the spoken answer; the full written answer is also in History. Your organisation does not create an Amazon skill.
Enable a private organisation address so trusted people can send or forward PDFs into the shared library — the same corpus as Upload. Manage trusted senders, regenerate the address, and review ingest activity from a dedicated Admin → Inbound Email screen.
Shape first-run onboarding: introduction routes, tenets, persona defaults, and Getting Started so new users land in the right research context.
Maintain the Question Bank — create, bulk edit, import/export CSV — to steer welcome prompts and suggested questions.
Manage Providers, Default Model (Ask and Virtual Review), Powerful Model (for retrying an unsatisfactory answer with a stronger option), Math Model, Voice / Alexa Model (spoken summary only; History keeps the full written answer), PDF Extract Model (Library document text and metadata), and which models are available. Bring Your Own Key (BYOK), optional web search per model, and Zero Data Retention where the provider supports it — without locking to a single vendor.
Compare how candidate models perform on your material before you widen access.
Set the chat attachment parse model and test extract behaviour on a sample PDF or image before that configuration is used for attachments in Ask.
Build an independent equation ground truth from Library PDFs, compare immutable extraction runs, and measure detection, transcription, and label quality — without changing what Ask retrieves.
Free Tier Settings, Tester Settings, Student and pay-as-you-go plans, subscription plans with feature flags (Personal documents, MCP Access, Web search, Alexa Access), active subscriptions, and revenue breakdown. Users can self-serve checkout, customer portal, cancel, and resume where enabled.
Cost Analytics covering document downloads, chat attachments, chat infographics, Q&A sharing (including social graph), MCP usage, answer ratings (satisfaction and downvote trends by model), and live presence of who is online.
Feedback Inbox — collect and triage user feedback from the in-product feedback button, plus answer thumbs up/down with optional comments, with export when you need a record. Open any rated answer to review the full conversation.
Generate an operational snapshot of the organisation’s configuration and content posture.
Server Settings and Client Configuration for exposing research capability to approved external agents and tools — including secure connect flows, keys, live connections, and tool controls (Beta). When personal documents are enabled for the caller, agents can list those documents and appraise them against the shared Library.
Install snippet, site key & origins, settings, and embed analytics. Visitors can ask from your site; handoff lets them continue the conversation after signing in to the full workspace.
Readers can open and download source PDFs from the Library; administrators see download analytics.
04 · User capabilities
Day-to-day, Quaestra.ai is a research companion. People work through the same named areas they see in the product navigation.
Ask natural-language questions with live streaming answers grounded in sources you can open and check. Paste formatted equations or an image of a formula into Ask — the math is captured into your question and shown in the conversation. Highlight any part of an answer (including equations) to open a quick explanation that keeps the underlying math, then copy it or continue in Ask as a new conversation that keeps the original turns and adds the explanation next. Optional Peer-Reviewed Only, Go Deep, author filters, follow-ups, and suggested questions when enabled. Ask stays on your library: pure off-library calculations are declined rather than computed. Rate each answer with a simple thumbs up or down so the organisation can see what is working. After marking an answer unhelpful, you can optionally try the same question again with a more powerful model when your organisation has enabled that option.
Turn an answer into a downloadable infographic image. Administrators choose the audience — keep it off, limit it to a preview group, or make it available to everyone in the workspace.
Revisit past conversations, rename or delete threads, continue multi-turn research, and export conversations as CSV, Excel, PDF, or Word.
Receive optional daily and weekly emails summarising your recent questions with short answer previews, including equations shown as formulas. One click opens the conversation in Ask after sign-in if needed. If you have been away with nothing new to summarise, you may still get a short come-back note with suggested questions ready to ask. Manage preferences in Profile, or unsubscribe from the email with a simple confirmation.
Where enabled by plan (and for admins and testers), upload personal papers for dual-scope research in Ask or via MCP Access, and run Virtual Review against the shared expert lens, with PDF and Word (DOCX) export of the review.
Persona preferences, daily and weekly Q&A digest controls, daily quota visibility, self-serve Subscribe / Cancel / Resume and billing portal, MCP Access integrations when on plan, Alexa Access to link the shared Quaestra.ai Alexa skill with a short pairing code (Assistant Name brands voice replies and lets users address their organisation inside the skill), voice Ask with a spoken summary (full written answer stays in History; math and sources on the Alexa card and History), Web search when a question asks for the internet (on plans that include it, and for admin/tester evaluation), and Delete Account.
Help, Getting Started, intro routes, and persona questions so new users understand how to work with the corpus and lens.
Report issues or share ideas through the feedback button, including optional file attachments, without leaving the workspace.
Share answers via controlled links or email. Recipients can open a public share, then Claim, Join, or Continue after sign-in. Copy for reuse, export as LaTeX, and download Word (DOCX) or PDF where the flow supports it.
Attach supporting files to a question when the organisation has chat attachments enabled — use the paperclip or drag a file onto the Ask box. Attached files are used as context for the answer and are not pasted into the visible question. Pasting an equation (or an equation image) into Ask is different: the math appears as a rendered block in your question that you can click to edit.
The research workspace is designed to work on phone, tablet, and desktop — including responsive navigation and touch-friendly Ask and History flows.
Register and sign in with passwordless email codes or single sign-on (Google / Microsoft) where enabled, with cookie consent and Terms / Privacy acceptance as required.
Bring your corpus. Keep your edge.