Your private document hub, fully explained.

What is LumenQube

LumenQube is a private, AI-powered document workspace with published installers for macOS, Windows and Linux. Instead of juggling separate apps, you get one connected family: LumenPDF, LumenSheets, LumenDocs, LumenDesign (boards & video), LumenSlides, LumenNotes, LumenApps, LumenFlow, and LumenForms—with LumenAgent at the centre to coordinate them from one message.

Every tool shares the same Library, familiar controls, focused AI assistant, autosave, sharing, history and account, so the experience remains one application. Core desktop editing is free to download and use without a trial timer. AI and some connected features draw on your allowance; every account gets a free monthly one, while an optional subscription replaces the need for separate app subscriptions when you need larger allowances or advanced connected capabilities. LumenQube is local-first: ordinary editing stays on your device, and content is sent to the cloud only for features you choose that require it (see Security & privacy).

What is included

Eleven editors — LumenPDF, LumenSheets, LumenDocs, LumenDesign and Video, LumenSlides, LumenNotes, LumenApps, LumenForms and LumenFlow — with LumenAgent at the centre to coordinate them from one message.

What it costs to use

Every editor is free, with no trial timer and no watermark. Opening, editing, exporting, redacting and keyword search never draw on an allowance. The app asks for a free account (no card) when you first open it. AI features and a few add-ons do — see AI in every editor & usage.

What “local-first” means here

Ordinary editing happens on your device and keeps working with the network off. Content is sent to the cloud only for the features you choose that require it, and only the part that feature needs. The full row-by-row breakdown is in Security & privacy — we would rather print the table than say “private” and let you assume the rest.

Platforms

Plus a mobile companion and a browser-based viewer for anything you share, neither of which needs an install.

Getting started

Download & install

Head to the download page — it detects your operating system and recommends the right build.

1. Download the build for your platform

   On macOS, take the .dmg for Apple Silicon (M-series) or the one for Intel — the download page picks the right one for you. On Windows, take the .exe installer for x64 or ARM64. On Linux, choose the .AppImage or the .deb.

2. Install it

   On macOS, open the .dmg and drag the app into Applications. For the .AppImage, mark the file executable and run it; the .deb installs through your package manager.

3. Open it once and allow it to launch

   If macOS blocks the app the first time, right-click it in Applications and choose Open, then confirm. This is only needed once — see Troubleshooting.

Create your account

On first launch you'll be invited to create a free account, or sign in if you already have one. You can sign up with an email and password (we'll email a 6-digit code to verify it) or continue with Google. New accounts include a free monthly AI allowance.

Your first job

From Home, message LumenAgent with the outcome you want, open an existing file, or create something new—a PDF, sheet, document, board, video, slide deck, notebook, app, automation, or form. Agent-created results open in the same full editors you use manually. Everything you create is saved automatically and appears in your Library for quick access later; a short welcome guide points out the essentials the first time you enter each tool.

The hub & library

The Home screen is your hub. From here you can:

• Start with the outcome — message LumenAgent, pick an editor, open a file, or choose a template; keep several documents open in tabs.
• Find anything — the unified Library searches across every file type and keeps a list of recent work. Group by type or folder, sort by date, name or size, and open a file once to keep it a click away.
• See your work — a "Your work" area surfaces what you've been doing across tools so you can pick up where you left off.
• Pick up shared work — the Shared section shows documents you've shared and documents others have shared with you.
• Continue in LumenAgent — Home search can hand its context to Agent, which can act across editors to create a document, transform a sheet, generate a deck, and more.

Your work auto-saves as you go, and the app keeps recovery copies so an unexpected shutdown won't lose your progress.

The Library

One Library covers every file type — documents, sheets, decks, PDFs, boards, videos, forms and notes — plus the pages you wrote. Group by type or folder, sort by date, name or size, and open a file once to keep it a click away.

Finding things

• Full-text search runs on your machine and works with the network off. It reads inside documents rather than matching filenames, so a hit reports the page or location it was found on.
• Searching by meaning — describing what you cannot name — needs an account and a connection, because the meaning-matching runs in the cloud. It is the one search that leaves your machine.
• Notes and files are one result set, not two tabs. The answer is usually in whichever half you would not have searched.

Spaces

A space is a place work lives. It owns no bytes, creates no folder on disk, and deleting one deletes nothing — it is a frame that changes what the Library, search and the agent show you by default.

• New work lands where you are standing, with no filing dialog. Derived work — a duplicate, an export — inherits from its source instead.
• A space shows what is filed in it, plus everything filed nowhere. Sharing carries filed members only, so loose documents appear in every space and are sent to nobody.
• Revoking a shared space closes every room it opened.

Tabs, recents and shared work

Several documents stay open in tabs. A “Your work” area surfaces what you have been doing across tools, and the Shared section shows both what you have shared and what others have shared with you.

Autosave and recovery

Work auto-saves as you go and the app keeps recovery copies, so an unexpected shutdown does not cost you your progress. Autosave and recovery reduce accidental loss — they are not a replacement for your own backups. See File lifecycle, history & recovery.

LumenAgent — your control centre

LumenAgent is the fastest way into the whole workspace. Give it an outcome instead of a list of clicks. It can plan the work, choose and coordinate LumenQube tools, read the sources you authorize, use connected services, create real deliverables, and show a live record of what it did.

What LumenAgent can do

• Plan multi-step work — break a goal into ordered steps, revise the plan as the job develops, and show each action and result on a live timeline.
• Find and understand source material — search your library, read documents and attachments, understand images, search the web, look at the active screen when needed, and work with files you select on your computer.
• Create and transform real work — build documents, spreadsheets, dashboards, reports, slide decks, boards, videos, forms, interactive LumenApps, and reusable automations; clean and transform source data along the way.
• Work across editors — delegate specialist work to the PDF, Document, Sheets, Design, Slides, Forms, Apps, and Automate tools instead of flattening everything into a chat response.
• Use computation when a task needs it — write and run Python in a sandbox after you approve the code-running step.
• Use connected services you choose — work with Google Drive, Gmail, Google Calendar, YouTube, OneDrive & SharePoint, Outlook, Dropbox and Slack under their individual permissions (Pro and above); add provider-hosted connections such as Linear, Notion, Sentry and Stripe when they are configured and available.
• Keep context — remember useful preferences and prior work across sessions, while letting you review and manage what the Agent is using.
• Work by voice or from mobile — dictate a goal, hear a response, continue an Agent conversation on the mobile companion, and attach photos, PDFs, or data files.

Write a useful request

A short message can be enough. For more control, include the five details below. You can add or correct any detail after the Agent proposes its plan.

• Outcome — what should exist when the job is finished: “a board-ready retention pack,” “a signed tenant packet,” or “a weekly feedback brief.”
• Sources — the files, folders, Library items, attachments, connected services, or public information it may use.
• Audience and standard — who will read or use the result, the tone, brand, level of detail, and any accuracy or accessibility requirements.
• Boundaries — what must stay unchanged, what may not be sent or published, a usage or step budget, and whether approvals should be strict.
• Destination — the output file type, folder, share setting, recipient, or recurring trigger. Leave this open when you want the Agent to recommend one.

Approvals, permissions, and recovery

• Permission presets decide which categories of work the Agent may attempt. A disabled capability stays blocked.
• Approval prompts appear before sensitive actions such as renaming, organizing, overwriting, deleting or changing automations, sending or publishing through a connector, and running code.
• Budgets let you bound steps and usage. The live timeline makes current work and completed actions visible.
• Cancel, retry, and recovery are available when a run needs correction. AI revisions create checkpoints, so you can review history and restore an earlier state.
• Manual control remains — open any output in its full editor, change it yourself, or continue with the editor-specific assistant.

See a visual walkthrough on the LumenAgent section of the home page. Connector-specific permissions and data handling are documented in Google integrations & data access and Microsoft, Dropbox & Slack connected-app access.

The four engines

Every LumenQube tool runs the same assistant. What an engine changes is not which assistant answers — it is how far that assistant goes before it calls the work finished: how many tool steps it gets in one turn, how many times it reads its own output back, and whether it re-derives the headline numbers a second, independent way.

Every run is the same five steps

The shape does not change between engines; the engine decides how many of these steps it gets and how many times the last two repeat.

Expert and Ultra loop steps 4 and 5 until the result stops changing. That loop is the whole difference between a confident first pass and an answer that has been argued with.

What each engine adds

Step budgets shown are LumenSheets’. Every editor sets its own, and a deeper engine always gets more room in all of them.

Which engine runs by default

• Free runs on Lumen Fast, and Automatic routing steps up to Core when a request needs it. Choosing an engine by hand starts on Plus. That default is deliberate: a free month should not be spent three times faster by somebody who never asked for more depth.
• Plus and Pro default to Core, Max to Expert, Ultra to Ultra.
• From Plus upward you can pin an engine per editor in Settings, so the sheet can sit on Expert while the notebook stays on Fast.
• The picker sits in every editor’s AI panel. Each editor takes every engine your plan includes.

Asking for an engine your plan does not include

The request is neither refused nor silently upgraded. It runs on the deepest engine your plan allows that is no deeper than the one requested — so a Pro account aiming at Ultra gets Expert, not the Pro default of Core. The server is authoritative; the desktop copy only decides what to show you, which is why editing the client cannot buy a deeper engine.

What each engine is for, per editor

The full comparison, with a live demonstration of the same job at each depth, is on the AI engines page.

AI in every editor & usage

LumenAgent coordinates work across the hub; each editor also has a focused AI panel that understands the document currently open. Depending on the surface, it can draft, rewrite, summarize, extract, redact, analyze and explain, create formulas, build charts and dashboards, generate layouts and images, design decks, transcribe audio, direct video, and more. You can attach files or images as context, choose response detail, review changes in place, and see your usage transparently.

What a credit is

All AI work is metered in credits, and a credit is priced from what the work actually costs to run: a request that reads more of your document, or runs on a deeper engine, uses more of them. A short question and answer costs little; a long document rewrite costs more; a generated video is by far the largest single draw in the app.

Your balance, and a breakdown by engine and by capability, is in Account → Usage. You can set an optional monthly budget in Account → Plan & billing to stop spending at a number you choose.

The monthly allowance

• Every account gets a free allowance of 300 credits, granted at the start of each calendar month. It does not roll over, and it is granted whether or not you also have a subscription.
• A subscription adds a monthly allowance on top, granted on each renewal: Plus 2,600 credits, Pro 7,500, Max 17,000 and Ultra 37,500. The figure in Account → Usage is the authoritative one for your account.
• A paid month is paced, not rationed. By default one calendar day can use up to 30% of that month’s allowance, so a single heavy day cannot empty the month. The free allowance has no daily limit. The app shows how much of today’s limit is left.
• Top-ups never expire. If you run out mid-month, a one-off top-up sits alongside your allowance and carries over indefinitely. See Account, plans & billing for sizes and prices.
• Running out does not lock the app. Every editor keeps working — opening, editing, exporting, redacting and keyword search never touch your allowance. Only the AI features stop until the next grant or a top-up.
• Credits have no cash value, are non-transferable, and are not refundable once spent.

What draws on your allowance, and what never does

The distinction is worth learning once, because it is what makes running out of credits an inconvenience rather than a lockout.

Keeping the bill predictable

• Set a monthly budget in Account → Plan & billing. Spending stops at the number you choose rather than at the number your plan allows.
• Cap a single run. A budget can be set on one agent run; at the ceiling it stops and reports rather than quietly continuing.
• Convert stable =AI() columns to values. An AI formula is charged on every recalculation, so it is best used as a one-off pass and then pasted as values.
• Match the engine to the job. Fast is the right answer to dozens of small asks; the deep engines earn their steps on work you would otherwise check by hand.
• Draft images before full quality. A draft render lands near the five-credit floor.

Attachments, references & vision

Most AI panels accept context you attach by hand rather than only the document you have open.

• Files and images as context — attach a photo and ask the assistant to describe or work from it, or attach a Library document as a reference for the one you are writing.
• Library references — point at a workbook, a note or last quarter’s version and the assistant writes from those rather than from a general impression of what such a document sounds like.
• Vision on its own output — several tools render what they made and look at it: LumenDesign proofs a board, LumenSlides inspects a rendered slide, LumenVideo watches the cut back. Problems it spots, it fixes.
• Screen context — the agent can look at the active screen when a task needs it, and only when you allow that capability.

Reviewing AI output

• Changes are shown in place, in the real document, so you review a document rather than a diff of a chat message.
• Every AI revision writes a checkpoint. Restore an earlier state from version history — see File lifecycle, history & recovery.
• A failed check is reported as a failure. When a reconciliation does not agree, or a cited quote is not found in the source, the result says so rather than presenting the claim with a lower confidence score.
• Nothing is sent that the feature does not need. An AI request carries the passages that request requires, not your library.

LumenPDF

Open any PDF and edit it like it was always editable — the page, fonts, links, annotations, bookmarks and form fields are preserved.

• Edit real text & images — change wording while fonts, spacing, and layout stay intact; place and adjust images.
• Organize pages — merge, split, reorder, rotate, crop, resize, compress, and lay out booklets or N-up prints.
• Forms & signatures — fill and create form fields; add signatures, dates and checkboxes, and sign with certificates.
• Secure redaction — permanently remove sensitive text and its underlying content, not just cover it; sanitize hidden metadata.
• OCR — make scanned PDFs searchable and editable, with AI-vision escalation for difficult scans.
• Review tools — highlight, comment, add shapes, links and attachments; compare versions and run accessibility checks.
• Convert — export a PDF to editable Word or Excel in one click, and protect files with a password.
• AI — summarize, extract information, answer questions grounded in the document, and redact sensitive data on request.

Learn more on the LumenPDF page.

What “editable” means here

Most tools export a new PDF that resembles the old one. LumenPDF edits the document you opened: click and retype on the page, and the fonts, spacing, links, bookmarks and existing form fields are all still in the file afterwards. Embedded images can be swapped, moved, resized or deleted in place.

Page work

Page handling is direct manipulation — no dialog asking you to type a page range. Add, delete, reorder and rotate by dragging; merge and split documents; crop, resize and compress; and lay out booklets or N-up prints for printing.

Forms and signing

• Fill an existing form — field types, tab order and validation from the original file are preserved.
• Build one — add text fields, checkboxes, dates and signature fields to a document that has none.
• Sign — add a signature, a date and initials, or sign with a certificate where the recipient requires one.

Redaction

Because redaction is destructive by design, keep the original until you have checked the result. See File lifecycle, history & recovery.

Scans, OCR and conversion

• OCR makes a scanned PDF searchable and editable. Difficult scans escalate to AI vision, which draws on your allowance; ordinary OCR does not.
• Phone photos of documents are accepted as input, not only clean scans.
• Convert a PDF to editable Word or Excel in one step. The result is an editable document rather than a re-print of the page.
• Protect a file with a password on export.

Review and comparison

Highlight, comment, and add shapes, links and attachments. Compare two revisions to see exactly what changed between them, and run an accessibility check before sending a document on.

Asking the document

The AI panel answers questions grounded in the document you have open — summarise it, extract specific information, or find and redact sensitive data on request. Ordinary editing, page work and keyword search never leave your machine; only the AI step does, and only for the passages it needs.

LumenSheets & dashboards

A fast, familiar grid with formulas, charts, Power BI-style dashboards, and AI built in.

• Formulas & functions — the calculations you expect, with live recalculation, whole-column references, a Name Box and Name Manager, Goal Seek, and Flash Fill.
• AI-native formulas — fill cells straight from a prompt with functions like =AI(), =AICLASSIFY() and =AIEXTRACT().
• Formatting & charts — style cells, apply conditional formatting and highlights, and build charts and KPIs.
• Dashboards — turn any range into a Power BI-style dashboard with charts, KPIs, matrix pivot tables, slicers, alignment tools and 16 themes; export to PNG and let AI explain the data.
• Import & export — bring in .xlsx, .xlsm, Excel templates, .csv and .tsv; large imports are handled carefully so you never silently lose rows. VBA is never executed, and macro-enabled inputs export safely as macro-free .xlsx. Restoring a native .lumensheet file is never truncated.
• Scale — the grid, editing, formulas and dashboards are built to stay responsive on very large workbooks.
• AI for data — clean, transform, summarize, profile and explain your data in place.
• LumenApps — turn a sheet into a focused interactive app backed by its data.

Learn more on the LumenSheets page.

Opening and importing

LumenSheets opens .xlsx, .xlsm, Excel templates, .csv and .tsv, and saves natively as .lumensheet. Import a file from the Library, from Home, or by dragging it onto the window.

• Macro-enabled files are safe to open. VBA is never executed. An .xlsm opens as data and exports as a macro-free .xlsx; the macros are not carried across, so keep the original if you still need them.
• Large imports are checked, not truncated. A big file is brought in with a row count you can verify rather than silently clipped, and restoring a native .lumensheet is never truncated.
• Delimiters and encodings are detected on CSV and TSV import; a file that guesses wrong can be re-imported with the delimiter set by hand.

Cleaning an export before you model it

The most common starting point is a system export where numbers are stored as text, rows are duplicated, keys are blank, and the same customer is spelled three ways. Ask the assistant to profile the sheet before it changes anything — it reads the data and reports what is wrong rather than guessing.

1. Profile it. Ask “what is wrong with this sheet?”. You get the count of type problems, duplicates and blank keys, found by reading the values rather than inferring from headers.
2. Retype and normalise. A whole column can be re-typed in one step; values can be split, merged, transposed or mapped.
3. Check the row count. De-duplication reports what it removed. Confirm the count matches what you expected before building on top of it.

Formulas, and why they are formulas

When the assistant builds a model it writes a formula into the cell rather than pasting a computed number, so the model still works when the data changes. You can inspect what feeds any cell and what depends on it.

• The functions you expect, with live recalculation, whole-column references, a Name Box and Name Manager, Goal Seek and Flash Fill.
• AI-native functions — =AI(), =AICLASSIFY() and =AIEXTRACT() fill cells from a prompt. These draw on your allowance every time they recalculate, so prefer them for a one-off pass and paste the values afterwards if the column is stable.
• Lineage — ask what a cell is built from and you get the chain, not a description of it.

It reads the numbers back

Dashboards

Any range can become a dashboard with charts, KPIs, matrix pivot tables, slicers and alignment tools, in any of 16 themes. Dashboards export to PNG, and the assistant can explain what the data is showing rather than only drawing it. Charts are chosen to fit the comparison being made and are labelled to be legible in both light and dark themes.

Limits and performance

• The grid, editing, formulas and dashboards are built to stay responsive on large sheets; performance depends on your machine and on how many volatile formulas a sheet contains.
• A sheet with many =AI() cells recalculates against your allowance. If a sheet is slow or expensive, convert stable AI columns to values.
• Everything except the AI features runs locally and works with the network off.

LumenDocs

A clean block editor with structure and AI in the margins.

• Rich formatting — headings, lists, checklists, tables, quotes, callouts, shapes, banners, and dividers, with drag-to-reorder blocks.
• Markdown & templates — type Markdown shortcuts to format as you go, or start from professional templates for letters, reports and memos.
• Accurate pagination — page-split guides show exactly how your document breaks across pages for printing.
• Comments & outline — leave comments and navigate long documents with find and an outline.
• Write with AI — draft, rewrite, summarize, restructure and adjust tone inline.
• Import & export — Word import/export, Markdown and HTML, and pixel-faithful PDF export.

Learn more on the LumenDocs page.

Writing from your own sources

Point the assistant at your notes, a workbook and last quarter’s version and it writes from those rather than from a general impression of what a report sounds like. It can produce sections, tables and content in one pass, and rewrite a single section without disturbing the rest.

Blocks

Type / anywhere to open the block menu. Beyond headings, lists, checklists, tables, quotes and dividers, LumenDocs carries designed blocks — covers, callouts, stat rows, pull quotes and editable diagrams — art-directed to the document’s own theme rather than dropped in as clip art. Blocks nest, and drag to reorder.

Figures wired to real cells

A chart in a document can be built from a real workbook rather than pasted as a picture, so the figure and the spreadsheet cannot disagree. See LumenSheets & dashboards for the workbook side.

The writing preflight

Pagination, comments and review

• Page-split guides show exactly how the document breaks across pages, so what you see is what prints.
• Comments and an outline make long documents navigable; find works across the whole file.
• Tracked changes and checkpoints — every AI revision creates a checkpoint you can restore. See File lifecycle, history & recovery.

Import and export

Word in and out, Markdown and HTML, and a pixel-faithful PDF export. Markdown shortcuts format as you type. Templates for letters, reports and memos are available when starting from nothing.

LumenDesign (boards)

An infinite, free-form design board for everything visual — diagrams, charts, social graphics, marketing, posters and print.

• Design freely — text, shapes, connectors, images, tables and a large library of recolorable elements, with alignment and grid-snap on an infinite canvas.
• AI design — generate whole layouts, diagrams, charts, palettes and vector art; auto-layout keeps elements tidy and overlap-free, and a vision "eyes" pass reviews and fixes quality.
• Templates & smart blocks — start from social posts, posters, cards, invitations and campaign boards, or drop in theme-aware smart blocks.
• Brand kit & effects — reuse brand colors and fonts; apply text effects and curved text, image masks and photo filters, and element animations.
• Magic resize & bulk create — adapt a design to new sizes or generate many variations at once.
• Present & export — a Present mode plus PNG, SVG and PDF export.

Learn more on the LumenDesign page.

It commits to one direction first

LumenDesign settles concept, palette, type pairing and depth language before anything is drawn, then composes the piece to that decision. That is why it hands you one considered result rather than nine variations you have to judge. Brand kits are applied at this stage, so the palette is yours rather than a generic one.

Purpose-built builders

A poster, a card, a logo and a carousel are different problems, and each has its own builder rather than one generic “make a design” call. Boards hold multiple artboards at their real output sizes.

It proofs its own work

Brand kits

A brand kit stores palettes, type and logos so every board starts in your own visual system. Business plans can mark a brand profile as approved for the whole workspace — see Family & Business plans.

LumenDesign — Video

Switch any LumenDesign into Video mode to build short-form video — made for Reels, Shorts and TikTok — without leaving the app.

• Timeline editing — a multi-track timeline with scenes and clips, transitions and element animation.
• AI direction — an AI cut director places hard cuts on the beat, adds match-cuts and a reactive color grade, and lands the edit on the drop.
• Captions & titles — kinetic captions and titles with 9:16 safe zones for vertical formats.
• Sound — a composed soundtrack across many genres with a large sound-effects library, plus a cast AI voice-over.
• Footage — drop in stock video from Pexels and Pixabay, or generate clips with AI (Kling · Veo).
• Export & publish — export MP4, WebM or GIF; publish straight to YouTube, or hand off to TikTok, Reels and Shorts.

Video lives inside LumenDesign — see the LumenDesign page.

The spine before the cut

Genre, edit style, look and script are settled first; a promo cut without a spine is a slideshow with music over it. If you have written a script, the edit follows it rather than improvising around it.

Format is decided early, and it is not a crop

16:9, 9:16 and 1:1 are chosen at the start and the shot is reframed for the format, not cropped from a landscape master. A vertical cut composed as a vertical cut is the difference between a usable reel and a letterboxed one.

Assembly, sound and captions

• Cut — scenes, footage, pacing and transitions in one pass, then the edit is snapped to the music so the cuts land where the track does.
• Sound — only motivated sound is spotted. The intent is a mix, not a library dump.
• Captions — word-timed, styled and burned in.

It watches the result back

The finished cut is reviewed and the problems it finds are fixed before you see it. As with every generated result, review it yourself before publishing.

LumenSlides

Build and present polished decks, from first draft to presenter view.

• Themes & layouts — consistent, professional slides with smart multi-column and card layouts, a saved-theme library and font pairings.
• Content — text, shapes, images, tables, charts, diagrams and video, with theme-aware smart blocks.
• AI deck generation — describe a deck and get a complete, structured draft; then refine any slide. A vision pass checks for overlaps and quality issues.
• Motion & delivery — slide transitions, element animations and click-to-reveal builds, plus a presenter view with speaker notes and a timer.
• Export — reliable PowerPoint (.pptx), PDF, and image export.

Learn more on the LumenSlides page.

The storyline is settled first

A deck fails in the outline, not the layout. The run locks what the room has to believe, settles audience, length and outcome, and gives every beat an action title — a claim somebody could disagree with, rather than a one-word label like “Revenue”.

Charts come out of your model

Figures are taken from the workbook, PDF or notes you pointed it at, so the deck and the model cannot disagree. Point it at the source rather than retyping numbers into slides.

It compares options and inspects the render

Where a choice matters — a cover, a layout for a difficult slide — two are built and judged against each other and the stronger is kept. The finished deck is then rendered and inspected before it is handed over.

Speaker notes and handoff

Speaker notes are written per slide, covering what to say and how to move to the next beat. Decks export as .pptx or PDF, or as a link — see Sharing, web viewer & guest portal.

LumenNotes

A OneNote-style space to capture and organize ideas.

• Notebooks → sections → pages — organize naturally, with fast full-text search.
• Rich notes — tables, callouts, checklists, highlights, and a slash (/) command menu.
• Wiki-links & to-dos — connect pages with [[links]] and round up every open task across notebooks.
• Audio & transcription — record voice notes and transcribe them to text.
• Encrypted at rest — sensitive notes and recordings can be encrypted using your operating system's secure keychain.

Learn more on the LumenNotes page.

Pages and structure

• Pages nest up to eight levels deep. A branch can be moved under another page from its context menu, not only dragged within its own list.
• Type / for the block menu — headings, lists, to-dos, tables, columns, toggles, equations, code and data boards.
• [[Wiki-links]] connect pages, and the page you linked to lists what points at it, carrying the sentence the link appeared in. A connection you made by hand can be walked from either end.
• A ticked box is the same commitment everywhere it appears. Tasks are tracked by identity rather than by their position in the document, so reordering a page never scrambles what is done.

Data boards — a spreadsheet living in the page

Type /data, pick a workbook and a sheet, and the note holds a live view of it. Nothing is copied. The block references the file, so when the sheet changes the page changes with it — and when you edit a cell in the note, the value is written back to the sheet.

• Ten views of the same rows — table, board, chart, cards, gallery, list, stats, calendar, timeline and pivot. Filter and group without touching the file.
• Edit from the note — change a cell, tick a box, or drag a card between lanes to change its value in the sheet.
• Saved views and row templates live on the board, so a board opens the way you left it.
• Open the workbook in LumenSheets whenever you want the whole thing.

Agent workloads

Type @ on any to-do — or any row of a data board — and pick a role. The agent runs in the background while you carry on writing, and writes what it found into its own card.

Workloads run on your machine. Quitting stops anything in flight, and anything still queued waits for you. A run that stopped part-way says so in its own card, and that card is part of the page — it is still there tomorrow.

Recording and transcription

Recording audio into a page is free and stays on your machine. Transcribing a saved recording uploads that audio to a managed service and needs Plus or above; long recordings are split into chunks first. The live transcript you see while recording comes from your own system’s speech engine. If you would rather nothing left the machine, record without transcribing.

Where notes live

The notebook is encrypted at rest on your own machine and there is no server copy of the body text. Notes are indexed alongside your documents, so one search box covers both — see The hub & library.

LumenApps

Turn an idea or a spreadsheet into a focused interactive tool without starting a separate software project.

• Build with AI — describe a tracker, calculator, dashboard, planner, or client portal and get a working app.
• Sheet-backed data — use a LumenSheets as the app's live data source, with two-way sync so edits stay connected.
• Components — tables, charts, forms, search and filters in a responsive browser layout.
• Preview & share — preview and refine, then publish a real interactive web link with controlled access — usable on mobile.

Building one

Describe the tool you want — a tracker, calculator, dashboard, planner or client portal — and you get a working app rather than a mockup. Refine it in a live preview until it does the job.

Sheet-backed data

An app can use a LumenSheets workbook as its live data source, with two-way sync: a change made in the app reaches the sheet, and a change in the sheet reaches the app. That is what lets an app be the front end for a spreadsheet somebody else maintains.

Components and layout

Tables, charts, forms, search and filters, laid out responsively so the result works on a phone as well as a desktop browser.

Publishing

Publish an app as a real interactive web link with controlled access. Access rules, expiry and revocation work the same way as everything else you share — see Sharing, web viewer & guest portal.

LumenForms & People

Create forms, polls, surveys and quizzes, publish them to the web, and collect real responses.

• Every question type — text, choices, ratings, dates and scales, plus photo and video statement blocks and editable field keys with answer piping.
• Real file uploads — collect documents and images as answers.
• Logic & style — branching rules, themes, and security gates.
• AI generation — draft a whole form from a one-line prompt.
• Publish & review — share a public or protected browser link under the access policy you configure; watch responses roll into a dashboard, see poll results, and export to LumenSheets.

LumenPeople (audience). Keep an audience of contacts and groups, email a published form to the right people, and resend to whoever hasn't responded yet — all from inside the app.

Learn more on the LumenForms page.

The parts of a form that are actually hard

Anyone can put four text boxes on a page. Quotas, scoring, calculated fields and routed outcomes are where a survey stops being a questionnaire and starts being a system — and those are the parts LumenForms builds for you when you describe what the form is for.

Field types

Choice, matrix, ranking, scale, slider, file upload and statement blocks, each with the options that make it useful rather than a generic input — picture choices with emoji or image cards, matrix questions with multi-select rows, and file uploads constrained by type, count and size.

Logic, piping and scoring

• Branching — skip, route and score based on an answer, so a respondent only sees the questions that apply to them.
• Piping — carry an earlier answer into a later question’s wording.
• Calculated fields — derive a value from the answers rather than asking for it.
• Routed outcomes — a different ending per path, rather than one thank-you page for everybody.

Quotas

Publishing and results

Building a form is free; publishing one to the web needs Plus or above. Responses come back into the app, where they can be summarised, exported to a sheet, or read as a protected form. See Sharing, web viewer & guest portal for access control.

LumenFlow

Turn repetitive, multi-step document and data work into reusable, agentic pipelines.

• Visual pipelines — chain steps on a canvas: ingest files or folders → extract, classify, clean and transform → merge, convert and build reports → export or pipe results into a spreadsheet.
• AI steps — include AI-processing steps that transform data and documents at scale.
• Save & reuse — save automation definitions and run them again on demand.
• Triggers & scheduling — run manually, on an interval or daily schedule, when a watched folder changes, when a form is submitted, or on a connector event.
• Reliable runs — a queue with concurrency control, an activity timeline, and safe retries with failure handling; runs continue in the background.

Learn more on the LumenFlow page.

The shape of a flow

A flow is a chain of steps — ingest, transform, export — where each step feeds the next. Drag steps into place and inspect or edit the pipeline visually; it is not a script you have to read to understand.

Triggers

Run a flow on a schedule (hourly, daily, monthly) or fire it on an event — a file arriving in a watched folder, a form response, or a connector event. Watching for a trigger costs nothing; you only spend allowance when one actually fires and the AI steps run.

What a flow can do

• Transform thousands of rows in the background rather than by hand.
• Tag, group and summarise unstructured input — support tickets, feedback, transcripts.
• Write a formatted report every run and deliver it as a file, a document, or rows piped straight into a spreadsheet.

Approvals still apply

A scheduled run obeys the same permission gates as an interactive one. Anything that sends, publishes or overwrites waits for you — a flow does not acquire permissions by being on a timer. See LumenAgent for the permission model.

LumenFlow and connected apps need Pro or above.

Search & findability

Four things make work findable here, and they are separate mechanisms rather than one feature with four names: full-text search over both halves of your workspace, hand-made links that survive being forgotten, tasks tracked by identity, and — for the thing you can describe but cannot name — search by meaning.

Search covers both halves

One search box runs over your notes and your library — documents, sheets, decks, PDFs and boards, not only the pages you wrote. It reads inside files rather than matching filenames, so a result reports the page or location the phrase was found on.

• It runs on this machine and works with the network off. The index is local; nothing is uploaded to answer a keyword query.
• Notes and files are one result set, not two tabs. The answer is usually in whichever half you would not have thought to search.
• Filter by kind — notes, docs, sheets, PDFs — after the fact, rather than choosing before you type.
• Home search and ⌘K are the same engine. One index, one result list; the two surfaces differ only in how much room they have to show it.

Links that hold

Write [[like this]] in a note to connect one page to another. The page you linked to keeps a ledger of what points at it, carrying the sentence the link appeared in — so a connection you made by hand can be walked from either end, months later, without remembering you made it.

A to-do list that agrees with your notes

Tick a box in a page and it is the same commitment everywhere it appears. Tasks are matched by identity, not by the line they sit on, so reordering a page never scrambles what is done, and the same task in a memo and in Today is one task rather than two that look alike.

Asking in your own words

Where the line is, for search specifically

More on the Search page, and the whole boundary in Security & privacy.

Calendar & your day

Google Calendar and Outlook merge into one list here, ordered by when things actually start. This is a view, not a second calendar: the providers keep the record, because they are better at holding it and because owning a schedule LumenQube does not control would be a promise it could not keep.

What the calendar view does

• Merges both providers into one fortnight, grouped by day, in your own time zone — including all-day events, which the two providers describe differently.
• Names the account on every row. Two calendars merge here, and a row you cannot place is a row you cannot act on.
• Opens any event in the calendar it came from, where changing it actually changes something.
• Lets you take a calendar — or one event — out of context. That is the part neither provider gives you, and it is the rest of this section.

What it deliberately does not do

• No month grid. A fortnight answers what is coming; a month is a filing cabinet.
• No drag-to-reschedule. Moving a meeting belongs where the invitations are sent from.
• No create-event form on this screen — though you can still ask the assistant to book something.
• It does not claim your provider forgets anything. Excluding a calendar changes what LumenQube may read. Google and Microsoft still hold the event.

Where “excluded” actually happens

Most software spells exclusion as a display filter: fetch everything, then hide the row. That is a preference, not a privacy control — the data still arrived. There are two controls here and they cut at two different moments, which is why they are two controls rather than one word.

1. Something needs your schedule.

   You open Today, ask the assistant about your week, or an automation comes due. All three arrive at the same place, because there is one path to a calendar in this app.

2. Per-calendar exclusion cuts first.

   An excluded calendar is removed before a request exists. Naming it explicitly does not get it back: an account hint can narrow what a call may touch and can never widen it, so asking about “my Outlook calendar” while Outlook is excluded returns nothing from Outlook rather than quietly re-admitting it.

3. The request goes out — to whatever survived.

   If nothing did, no request is made at all, and the assistant is told your calendars are excluded rather than telling you to connect a calendar you can see is already connected.

4. Per-event exclusion is a filter, and we would rather say so.

   A single event can only be recognised once the list has arrived, so an excluded event is fetched and then discarded before display. A whole calendar can be stopped before the request; one event inside it cannot.

5. What is left is what everything sees.

   Today, the calendar screen, search and the assistant all read the same filtered list. There is no second route that reaches a calendar without passing through both cuts.

What exclusion reaches

A control you can see, and undo

• Each calendar says what its state means for the app, not for the screen — “LumenQube does not read this calendar, and the assistant cannot see it” rather than a switch position you have to interpret.
• Excluded events stay listed, by id, with no titles. Keeping the title would mean keeping a copy of the thing you hid. The id is enough to put it back.
• “Nothing on” and “all excluded” are different screens. Being told your fortnight is clear seconds after hiding it yourself reads as a broken control rather than a quiet week.
• Including a calendar again re-fetches it, because nothing was kept. There is no cached copy sitting behind the control.

Exclusion is documented end to end on the Calendar page; connector permissions are in Connectors.

Meetings

A meeting used to be four unrelated records — the invitation, the people, the hour and the recording. LumenQube keeps the meeting itself and hangs the rest off it, so a note linked to an event resolves to the same meeting as the event does.

What a meeting holds

• One identity from both directions. The appointment and the recording of the appointment are one thing rather than two that look alike.
• Who was actually there. Both calendars always sent the guest list and both used to throw it away and keep the count. The identities come through now.
• The transcript, with turns and timings — and the decisions and action items drawn out of it, each carrying a quote.
• It is registered as sensitive, so it stays out of the plaintext index alongside your notes and tasks.
• A 400-person all-hands truncates its list and never loses its number. Attendees are capped on the wire and again on arrival, with the count kept separately.

A citation you can check, or no citation at all

Most summarisers report a citation: the model says where it got something and you take that on faith. This one verifies it, in three steps.

1. Proposed.

   The model reads the transcript and returns a claim plus the words it says were spoken — an actual quote, not a paraphrase and not a timestamp.

2. Looked up.

   LumenQube searches the transcript for those words — not the model’s account of where they were — and reads the time off wherever they actually landed.

3. What reaches you.

   A claim whose quote is not in the transcript is discarded and counted. Not shown with a warning, not shown with a lower confidence score.

Where it appears

The write-up sits under the recordings strip on the note, collapsed by default, below your own writing — which is unchanged. A write-up that rearranges your own notes is not a feature; your page is byte-identical after extraction. Action items become real tasks.

What is not built

• No speaker identification. This does not diarise. Where the source labels speakers they are kept; where it does not, a turn says “speaker not identified” rather than guessing a name.
• No person page. “Who was in that meeting” is answerable because the attendance links exist, but a surface showing one person’s meetings and commitments is not built.
• Nothing runs while the laptop is shut. Meeting artifacts are gathered by this app, on this machine, when it is awake.
• No Zoom or Teams. The calendars are Google and Outlook. LumenQube does not join your meetings as a bot.

More on the Meetings page.

Memory & context

There are two kinds of memory here, they are stored differently, they travel differently, and they are forgotten differently. LumenAgent settings → Memory (also reached from Settings → AI) shows both, shows the exact text they will add to your next question, and erases either one.

Durable memory — what you told it

• Facts the assistant saved because you said them — preferences, projects, people, goals.
• Written only when something is durable. Credentials, keys, card numbers and recovery phrases are screened out before anything is saved or synced.
• It rides the cached part of the request, sent once rather than re-sent with every question.
• Switch it off and the block is not built at all. There is no second copy elsewhere in the request.
• Memories carry no source links. You can read what was remembered, not the conversation it came from.

The context graph — what it connected

• Links between your own notes, tasks, meetings, events, people and files. Nothing is fetched to build it; it is drawn from what you already have.
• It lives on this computer, in a local database beside your documents — not synced, not backed up to us, not readable by anyone who has not unlocked this machine.
• It travels as titles and distances, never contents. The context block carries what you are working on and what is related to it, and how many steps away — not the text of any of it.
• Notes and tasks are held as sensitive and stay out of the plaintext index entirely, which is why the graph can reach an encrypted note that no search query could.

Forget, and the five places it has to reach

Deleting a note is one row in one table. Forgetting it is five operations across four stores, and no cascade spans them — which is the whole reason this is a feature rather than a delete button.

What this screen does not do

• No per-space or per-project AI-access matrix. You cannot yet say “this project is never readable”. The control does not exist, so the words for it are kept off the screen.
• Home’s quick search does not use the graph. It stays one round trip on purpose. The ⌘K library search and the assistant both do use it.

The exact context block, printed verbatim, and the counts by type are on the Memory & context page.

File formats & exports

LumenQube works with familiar office, data, image, and video formats while using native LumenQube formats to preserve the richest editable state. Availability depends on the editor and whether a format is being opened, imported, exported, or published.

How to read this table

Availability depends on the editor and on whether a format is being opened, imported, exported or published. Native LumenQube formats preserve the richest editable state; an office format preserves what that format is capable of holding.

Notes on specific formats

• Macro-enabled workbooks (.xlsm) open as data. VBA is never executed, and the export is a macro-free .xlsx — keep the original if you still need the macros.
• Scanned PDFs and phone photos are accepted as PDF input; OCR makes them searchable and editable. See LumenPDF.
• PDF export is pixel-faithful from LumenDocs — what the page-break guides showed is what prints.
• A native file is never truncated on restore. Large imports report the row count rather than silently clipping.
• Fonts are embedded on export where the format supports it, so a document opened elsewhere does not re-flow.

Choosing between native and interchange

Work in the native format while the work is live — it is the only one that keeps every editable property. Export to an office format at the point of handover, and to PDF when the recipient should not be editing it at all. If a document will come back to you for another round, ask for the native file or the office format, not the PDF.

Sharing, web viewer & guest portal

Use the Share button in any editor to collaborate or publish.

• Share & collaborate — invite people by email as viewer, commenter or editor, or with a revocable private link. Comments, version history, live presence and an activity feed stay attached to the work, and conflicting edits are merged safely rather than lost. A "Shared with me" inbox and email notifications keep everyone in sync.
• Publish to the web — create a view-only web link that mirrors the desktop view, at view.lumenqube.com. It works for PDFs, documents, sheets, boards, slides, interactive dashboards, LumenApps, and forms & polls. Choose public, restricted or passcode-protected access, optionally allow downloads, and revoke the link anytime.
• Guest portal — guests and collaborators get a friendly library of everything shared with them instead of a pile of links: search, filter, sort, favorite and switch layouts; see forms and polls that need a response; review access, owner and protection status; and move to a full desktop account whenever they're ready.
• You stay in control — shared and published content is encrypted at rest on our servers, you can change roles or rotate a document's key, and you decide who has access.

The three ways to hand something over

Controlling a share link

• Expiry — a link can stop working on a date rather than living forever.
• Downloads — allow or withhold them separately from viewing.
• Revocation — withdrawing access closes the room. A person removed from one document stays removed.
• Sharing a space carries only the documents actually filed in it — see The hub & library.

What a shared document is, technically

The guest portal

A recipient does not need an account or the desktop app. The web viewer renders the document faithfully in a browser, including on a phone, and respects the access rules attached to the link.

Mobile companion

LumenQube is an installable, phone-first companion—not a cramped copy of every desktop editor. It keeps LumenAgent, shared work, tasks, and safe desktop handoffs available when you are away from your computer.

• Continue Agent conversations — use quick, tool-free cloud chat on the phone, or hand richer work to a paired, signed-in desktop and see the answer return to the same conversation.
• Bring mobile context — attach up to four selected photos, PDFs, or supported text/data files to a quick-chat turn; attachments are bounded and processed only for that turn.
• Find and open work — search a bounded metadata index of recent desktop work, shared documents, owners, projects, forms, dashboards, and LumenApps, then open browser-friendly shared items on the phone.
• Track handoffs — review queued and running tasks, stop an in-flight response, cancel a queued handoff, and keep new requests in an encrypted offline queue until connectivity returns.
• Manage conversations and access — pin or delete synced conversations, inspect signed-in devices, revoke sessions, recover or rotate a password, and wipe mobile sync data.
• Install to the Home Screen — the mobile web companion runs as a standalone PWA with a cached static shell; APIs, account data, AI streams, and documents are not cached by the service worker.

What the companion is for

The mobile companion is a second surface on the same account, not a second product. It is for the moments away from the desk: capture something, ask a question, check on a run, and continue a conversation you started on the desktop.

What you can do from it

• Continue an agent conversation started on the desktop, and start a new one.
• Attach photos, PDFs and data files from the phone as sources.
• Dictate a goal and hear a response.
• Read your library and open documents in the mobile viewer.

Where the work actually runs

What stays on the desktop

The full editors are a desktop experience. The companion is for capture, questions and continuation; editing a spreadsheet model or laying out a deck is not something it tries to replace.

Connectors

Connect outside services so the assistant, data tools, and publishing workflows can work with them. Each connection starts through the provider's consent screen, and access tokens are encrypted in a server-side vault — LumenQube never sees your provider password.

• Google Drive — search and read existing files, import and refresh a Google Sheet or CSV, and create a new Google Doc after you approve the action.
• Gmail — search and read messages for a task; send a plain-text or safely formatted message or reply only after you approve the action.
• Google Calendar — list and search events; create an event or RSVP at your direction.
• YouTube — upload only the finished LumenDesign video you choose, using the title, metadata, audience, and visibility you review.
• Outlook — search and read mail, safely send formatted or plain-text mail and replies, work with calendars, and look up contacts. Consequential actions require your approval.
• OneDrive & SharePoint — search and import files with read access. User-directed save-back can request incremental write scopes; the default connection remains read-only.
• Dropbox — search, import, and refresh selected file data with read-only file permissions.
• Slack — list and read only conversations joined by the LumenQube bot, search recent joined-conversation history locally, and post only after approval.
• Manage anytime — disconnecting deletes LumenQube's stored authorization for that service. You can also revoke access from the provider's account controls.

See every requested permission, its exact purpose, data handling, and user controls on the Google integrations & data access and Microsoft, Dropbox & Slack connected-app access pages.

The two kinds

Connectable services fall into two kinds, and the difference matters for what you are trusting.

• Eight first-party connectors, built and maintained here, each with its own OAuth consent screen: Google Drive, Gmail, Google Calendar, YouTube, Microsoft OneDrive, Outlook, Dropbox and Slack.
• Additional provider-hosted connections — Linear, Notion, Sentry, Stripe and others — are reached through their own published servers when they are configured and available.

Granting and revoking

• Every service is off until you connect it individually. Connecting one grants nothing to the others.
• Only the permissions the feature needs are requested at the consent screen. You can read exactly what is being asked for before you approve it.
• Tokens are held in a server-side vault, never written to disk in the clear.
• Disconnect at any time. This deletes LumenQube's stored authorization record and asks the provider to revoke its grant where the provider offers that: Google once no other Google service you connected relies on it, Dropbox, Slack user tokens, and hub apps that publish revocation. Microsoft access and Slack workspace installs are revoked in the provider's own account controls. Completed work and saved or remote content are unaffected.

What the agent may do with a connection

A granted connection makes a service reachable; it does not make it writable without asking. Reading is routine. Sending a message, publishing, or overwriting something in a connected service stops and waits for your approval — on every permission preset, and on scheduled runs too. See LumenAgent.

Calendar exclusion

Google services specifically

Connector-specific permissions and data handling for Google services are documented in Google integrations & data access. The full list of connectable apps is on Connected apps.

Account, plans & billing

A free account (no card) is required to use the app. Signing in is how it knows your plan and allowance; ordinary editing still happens on your device and your documents are not uploaded. AI, sharing, publishing and the mobile companion use the account’s cloud features.

The plans

Prices below are the USD basis. You are charged in Canadian dollars, converted from that basis — the figure shown at checkout is what your card is billed and is the authoritative one.

Paying annually charges ten months for twelve — two months free, about 17% off. Family plans are one flat monthly price for the household, and Team and Business plans are priced per seat; both are covered in Family & Business plans.

Top-ups

A top-up is a one-off purchase for when a subscription month runs short. Top-up credits never expire, and they are spent only after the month’s allowance is gone.

Managing the subscription

• Subscriptions renew automatically and can be cancelled at any time. Access continues to the end of the period you have already paid for; it is not cut off at the moment you cancel.
• Changing plan takes effect on the billing cycle, and your monthly allowance changes with it. Top-up credits are unaffected in either direction.
• Usage — Account → Usage shows what is left, by day, editor and engine; top-ups, auto top-up and the optional monthly budget are in Account → Plan & billing.

Profile, sign-in and devices

• Profile & settings — manage your name and email in Account → Profile, and preferences in Settings.
• Password — reset it from the sign-in screen via Forgot password?; a 6-digit code is emailed to you.
• Signed in with Google? That account has no password, so the change-password form does not apply to it and the identity is managed in your Google account. Accounts created either way behave identically everywhere else.
• Devices — see every device signed in to the account, and sign others out.

Billing questions and refunds

Contact support@lumenqube.com for anything billing-related, and see the refund terms for what is refundable. Credits already spent are not refundable, and credits have no cash value.

Family & Business plans

Family and Business plans let several people share one subscription. Business adds organization branding and server-enforced governance; see the Business overview.

• Pooled allowance — the plan's monthly AI allowance is shared across everyone on it.
• Members & invites — the plan owner invites members by email and manages seats.
• Roles — members can be owners, admins or members. Business owners decide whether admins may invite people, manage members and limits, or manage member devices; administrator roles stay owner-controlled.
• Business brand — maintain an approved raster logo, semantic colors and heading/body typography, then any member can add that profile to their own Brand Kit in LumenDesign with one click.
• Domain and link policy — restrict seat and document invitations by email domain, disable new public links, or require every new share link to expire.
• Administration activity — owners and admins can review a bounded, redacted trail of recent workspace changes; document content, credentials, IP addresses and device identifiers are not included.
• Per-device sessions — each member signs in on their own devices and manages their own sessions.

Manage a plan group from your Account page, or see pricing for what's included.

What a Business workspace adds

• One pooled AI allowance for the organisation rather than a separate one per person.
• An approved brand profile — the palette, type and logos every board, deck and document starts from. See LumenDesign.
• Role-based administration, domain controls and share-link policy.
• Member and device controls, including removing access for somebody who has left.

Spending ceilings

AI spending can be bounded at more than one level, and the ceiling that applies to a run is the narrowest of the ones above it: the workspace, then the space, then the assignment, then the run itself. A tighter ceiling below never widens a looser one above.

Unattended work

Whether members may run work that continues while they are away is a workspace setting, not an individual one. It is enforced on the server for the same reason.

Seats and billing

Business plans are priced per seat with a minimum seat count, billed on the same cycle as an individual plan. See the Business page for current pricing and the Account, plans & billing section for how billing works generally.

Security & privacy

LumenQube is built local-first. Your documents stay on your device by default and aren't uploaded in the background. Content is transmitted only for the cloud features you choose to use — AI, sharing, web publishing and connectors — and only the content that feature needs.

• Data in transit is encrypted with TLS; shared and published documents are encrypted at rest on our servers with server-managed keys. They are not end-to-end encrypted.
• Sensitive local data such as notes and audio can be encrypted at rest using your operating system's secure keychain.
• Connector tokens are held in a server-side vault; you can disconnect a service at any time.
• LumenQube does not use customer content as generalized model-training material. A provider's own data use, retention and permitted security or legal processing depend on the service and our account terms and settings.

Read the practical control overview in our Security Centre and the data-flow, subprocessor, retention, and rights details in our Privacy Policy.

Where the line is

“Local-first” is worth being precise about. Every row below is one thing you do, and whether it crosses between your machine and the network.

What we do not claim

AI providers and training

LumenQube uses commercial AI services through their APIs and does not use customer content as generalized model-training material. A provider's own data use, retention and permitted security or legal processing depend on the service and our account terms and settings. Request data may also remain in product state for the feature the user chose; see the Privacy Policy.

Connectors and tokens

Connector authorisation tokens are held in a server-side vault, never on disk in the clear. Every service is off until you connect it individually. Disconnecting deletes LumenQube's stored authorization record and asks the provider to revoke its grant where the provider offers that: Google once no other Google service you connected relies on it, Dropbox, Slack user tokens, and hub apps that publish revocation. Microsoft access and Slack workspace installs are revoked in the provider's own account controls. Neither step undoes completed actions or deletes saved or remote content. See Connectors.

What you control

• Per-capability permissions for LumenAgent — reading, writing, sending and code-running are separate decisions. See LumenAgent.
• Calendar exclusion — a whole calendar or a single event can be excluded from context. An excluded calendar is dropped before the connector is called, so no request is made for it at all. That is not the same as hiding it from a screen.
• Memory — what the assistant has remembered can be reviewed and erased.
• Device list — see everything signed in and sign others out.

The practical control overview lives on the Security & privacy page, and the legal terms in the Privacy Policy.

Settings reference

Settings is one window with a search field at the top of it — type what you are looking for rather than hunting the tree. It has nine sections, listed below in the order they appear, so you know what exists.

The nine sections

Outside Settings

Two things you might look for here live elsewhere. Account — overview, usage, plan & billing, profile, sign-in & devices, and your data — has its own page; see Account, plans & billing. Memory lives in LumenAgent settings → Memory, reached from Settings → AI; see Memory & context.

Updating the app

On macOS and Windows, LumenQube updates itself in place. Signed production builds download and install over the air; on Linux, and for an unsigned or ad-hoc build, or a temporarily unavailable update feed, falls back to the published installer, which you can always fetch yourself from the download page.

What the states mean

Required updates

Most updates are optional and you can keep working indefinitely without taking one. A release can be marked required for security or compatibility, and in that case you are prompted to update before continuing.

Integrity

Every downloaded payload is verified before it can be installed — the update must match the version it claims to be and the digest published for it. A download that fails verification is discarded rather than offered to you.

If an update will not install

• Nothing happens when you press restart — the update is still preparing. Give it a minute; the state changes to Ready to restart on its own.
• The update repeatedly fails — download the current build from the download page and install it over the top. Your documents, Library and settings are untouched by a reinstall.
• macOS refuses to open the new build — see Troubleshooting for the first-launch security prompt.

Keyboard shortcuts

On macOS the modifier is ⌘; on Windows and Linux it is Ctrl. Below, Mod means whichever applies on your platform.

Anywhere in the app

Home, the Library and the assistant

LumenPDF

LumenSheets

The grid follows Excel’s bindings wherever Excel has one, including the two that are muscle memory for anyone who has kept a spreadsheet.

LumenSlides

LumenDesign

Copy, cut and paste on the board run through the native clipboard, so pictures and text from other applications come in directly and an external copy is never shadowed by a stale in-app one.

LumenDocs and LumenNotes

LumenDesign — Video

Troubleshooting

The app won't open on macOS

If macOS blocks the app on first launch, right-click the app in Applications and choose Open, then confirm. This only needs to be done once.

AI features are unavailable

AI requires a signed-in account and some allowance left. Check that you're signed in (Settings) and that your allowance isn't used up. On a paid plan, also check whether today’s limit (30% of the month) is used up. If you're out, your free allowance refreshes at the start of each month, and on Plus and above you can buy a top-up.

I didn't get my verification or reset email

Check your spam folder and confirm you used the right address. You can request a new code from the sign-in screen. Still stuck? Email support@lumenqube.com.

A large spreadsheet import looks incomplete

Very large files may be capped based on your device's memory to keep the app responsive; when that happens the app tells you exactly what was left out. Restoring a native .lumensheet file is never truncated.

A connected service stopped working

Open Settings and reconnect the service — access can expire or be revoked on the provider's side. Reauthorizing restores it without affecting your documents.

An import or export appears stuck

Keep the app open and check the progress indicator before starting the same job again. Large PDFs, media projects, presentations, and workbooks can spend time preparing fonts, images, formulas, or video frames after the progress bar first appears. If progress does not change, cancel once, save a recovery copy, verify that the destination has enough free space, and retry with the original source. For an export, choose a new filename so an incomplete file is not mistaken for the final version.

The exported layout or font looks different

Confirm that the source font is installed and that substituted fonts have not changed line wrapping. Recheck page size, margins, bleed, slide dimensions, and responsive app breakpoints. For a document going to another device, PDF is the most dependable review format; keep the native LumenQube file for later edits. When exact typography matters, inspect the exported file at 100% zoom before delivery.

A shared link is unavailable

Open the item's Share panel and confirm that publishing is still enabled, the link has not expired, and the visitor meets any password or account requirement. If you changed visibility, republish and copy the newest link. Workspace administrators may also restrict public links. Disabling a link takes the published copy offline without deleting your local working file.

A video export is taking longer than expected

Rendering time depends on duration, resolution, frame rate, effects, captions, and source media. Keep source files connected, avoid moving them during export, and close memory-heavy applications if your device is under pressure. Test a short range at the intended settings before committing to a long final render.

I need to recover after a crash or forced restart

Reopen LumenQube and check Home, Recent, and the file's history before importing the original again. Autosave may restore a newer working state or offer a recovery copy. Save the recovered version under a new name, compare it with the last known-good export, and only then replace the previous working file. If recovery does not appear, include the file type, approximate time, and last completed action in your support report.

Form responses or automation results are missing

Check that the form is published, the automation is enabled, and its trigger is listening to the current form or folder rather than an older copy. Open the run history to see whether the job never started, paused for approval, or failed at a connected step. Reconnect an expired service, correct the failed step, and rerun only the affected item to avoid duplicate messages or records.

A mobile handoff is queued

Keep both devices online and signed into the same account, then reopen the transfer from the mobile companion. Large scans and recordings may wait for a stable connection. A queued handoff does not replace the source on your phone; once it arrives, verify the page order and file name in the desktop Library.

The web guide or viewer shows an older version

Refresh the page, then confirm that you republished after the last edit. If the old version remains, open the link in a private window to rule out a cached page. Published copies and local drafts are separate states, so saving the desktop file does not silently change what external visitors can see.

Something went wrong

Open the command palette with Mod ⇧ P and choose Report a bug…, or use the Help button. Recent diagnostic logs are attached automatically, which is usually the difference between a fixable report and an unreproducible one.

End-to-end recipes

These recipes show how LumenAgent coordinates multiple tools while keeping the files, checkpoints, and approval decisions visible. Use them as starting points rather than rigid templates: change the audience, output, review rules, and destinations in your message.

Write a strong Agent request

A useful request names the outcome, sources, audience, constraints, review gates, and deliverable. For example: “Using the signed statements in this folder, reconcile the Q2 totals against the approved workbook, explain differences over 5%, update the board deck in our existing style, and stop for my approval before publishing a PDF.” Agent can infer the tool route, but it should not have to guess the business rule or approval boundary.

File lifecycle, history & recovery

A LumenQube file moves through distinct states: source, editable working file, saved revision, exported deliverable, and—only when you choose—shared or published copy. Keeping those states clear makes collaboration safer and recovery faster.

Autosave, versions, and backups are different

• Autosave protects recent work in the active file. Wait for it to complete before closing the app or ejecting storage.
• History or recovery helps return to an earlier working state. Availability and detail vary by tool and file type.
• Exported copies preserve a delivery state but may not contain every editable feature of the native project.
• Device or organizational backups protect against device loss, storage failure, or broader deletion. LumenQube does not replace your backup policy.

Safe recovery after an unwanted AI or automation change

1. Pause the active Agent job or automation so it cannot repeat the change.
2. Duplicate the current file or save a recovery copy before restoring anything.
3. Inspect the action timeline or run history to identify the affected files and connected destinations.
4. Restore the most recent known-good state, then compare content, formulas, layout, permissions, and published outputs.
5. Correct the instruction, input, or workflow condition and test it on a copy before resuming.

Accessibility & inclusive output

Accessible work starts in the editable source, not at the final export. LumenQube provides keyboard-friendly controls and tools that help you create clear documents, but the author remains responsible for reviewing structure, reading order, contrast, descriptions, captions, and the exported result.

Final accessibility review

1. Navigate the complete output with the keyboard only and confirm focus remains visible and logical.
2. Zoom to 200% and check that content remains readable without losing controls or essential information.
3. Verify headings, lists, tables, labels, reading order, link purpose, and document language.
4. Check contrast and ensure color, motion, audio, or layout is never the only carrier of meaning.
5. Review alt text, captions, transcripts, and error messages for accuracy rather than mere presence.
6. Open the actual exported or published version—the editable source can behave differently from the deliverable.

Glossary

The words this guide uses in a specific sense, rather than a general one.

Agent

The assistant when it is given an outcome rather than an instruction — it plans, chooses tools, works across editors and reports what it did. Distinct from an editor’s AI panel, which works inside one open document. See LumenAgent.

Approval

A stop before a consequential action — sending, publishing, overwriting, deleting, running code. Approvals apply on every permission preset and on scheduled runs; a flow does not acquire permissions by being on a timer.

Brand kit

A saved palette, typography and logo set that boards, decks and documents start from. On Business plans it is an approved organisation profile rather than a personal one.

Checkpoint

An automatic restore point written before every AI revision. Distinct from autosave (continuous, unnamed) and from a backup (yours, off-machine). See File lifecycle, history & recovery.

Connector

An authorised link to an outside service. Eight are first-party with their own consent screens; additional provider-hosted connections may be offered over the open connector standard. See Connectors.

Context graph

Links between your own notes, tasks, meetings, events, people and files, built locally from what you already have. It travels as titles and distances, never contents. See Memory & context.

Credit

The unit AI work is metered in, priced from what the work actually costs to run — a deeper engine or a longer document uses more. See AI in every editor & usage.

Data board

A live view of a workbook sheet living inside a note. Nothing is copied — the block references the file, and editing a cell in the note writes back to the sheet. See LumenNotes.

Durable memory

Facts the assistant saved because you told it — preferences, projects, people. Separate from the context graph, stored separately, and forgotten separately. The desktop set and the phone set do not reconcile.

Engine

How far the assistant goes before calling work finished: Fast, Core, Expert, Ultra. A deeper engine takes more steps, and each step is priced from what it costs to run. See The four engines.

Exclusion

Taking a calendar or a single event out of what the workspace may read. A calendar is dropped before the request exists; an event is dropped after the list arrives. See Calendar & your day.

Handoff

Work started on the mobile companion and dispatched to a paired desktop, or to the cloud when no desktop is present. Where a run executes changes which spending ceilings apply.

Local-first

Ordinary editing happens on your device and keeps working with the network off. Content crosses only for features that require it, and only the part that feature needs. The row-by-row table is in Security & privacy.

Native format

A LumenQube file type — .lumendoc, .lumensheet, .lumenslides, .lumendesign, .lumenform — that preserves the richest editable state. An office format preserves what that format can hold.

Preflight

A review pass before a document leaves you: hedges, unsupported claims and passages asserting more than the sources carry, reported rather than silently reworded.

Quota

A server-enforced limit on how many responses a form accepts. Not a client-side check a determined respondent can walk past. See LumenForms & People.

Redaction

Removing matched content and the layer beneath it, then sanitising hidden metadata — not drawing a black rectangle over text that stays selectable underneath. Destructive by design. See LumenPDF.

Space

A frame that changes what the Library, search and the agent show by default. It owns no bytes, creates no folder on disk, and deleting one deletes nothing. Sharing a space carries its filed members only.

Top-up

A one-off credit purchase that never expires and is spent only after the month’s allowance is gone.

Workload

A background agent assigned to a single to-do or data-board row in a note, in one of six roles. Its output accumulates in its own card and is written into the page exactly once, when it settles. See LumenNotes.

Getting help

Before writing in, two things resolve most questions faster than an email: the search box at the top of this guide covers every section on this page, and Troubleshooting covers the failures people actually hit.

Reporting a bug from inside the app

Open the command palette with Mod ⇧ P and choose Report a bug…, or use the Help button. The report attaches details about your build and recent errors automatically, which is usually the difference between a fixable report and an unreproducible one. If support asks what the app was doing, Mod ⇧ D opens the in-app log console.

Who to write to

What to include

A report that names the editor, what you expected, what happened instead, and whether it happens every time is usually enough to reproduce. If a specific file is involved, say so — but do not attach confidential material unless support asks for it and tells you how to send it.

Elsewhere

Support & FAQ · Security & privacy · Privacy Policy · Terms of Service · Guides and walkthroughs