Privacy policy

Last updated October 5, 2026

This describes what the app stores about you, who else sees any of it, which services process it for us, and how to have it deleted. The app is a personal reading tool: nothing here is used to build an ad profile, and nothing is sold.

What we store

• Your email address, used to sign you in — and, if you sign in with Google, the name and picture Google gives us.
• The books you upload, the text extracted from them, and everything generated from that text or file — pictures, music, chapter overviews, recaps and so on — kept under a storage location tied only to your account.
• Your place in each book, the furthest you’ve read, and which days you’ve read.
• Highlights, notes, margin reactions, your diary, and the day’s question when you keep it.
• Your profile (name, bio, picture, banner, Four Favourites, recommendation sets), your Want to Read, your friendships, and the devices you’ve connected to the app (kept only as a scrambled token).
• Posts you make in the Agora, notes you leave for the next reader, and posts in a book you read together with friends — each kept with an internal record of who wrote it (for enforcing the rules and handling reports) — and any reports you make or blocks you set. A thread that leaves its board stays in that board’s archive, read-only, for 17 days, and is then deleted — unless a report holds it for review.
• When you confirmed you’re 18 or older, for the Agora, on your own profile.
• Quotes you pass to a friend, with the line you add, and the notes friends write back to you — each kept for the one friend it’s for until they put it away, and gone when you make the quote private again or the friendship ends.
• If the Agora has paused you from posting, a record of the rule, when it started and when it ends — kept with your account, and gone with it.
• If a post of yours was removed under a rule, a private note saying which rule and where (its words are read back from the post, never copied), and, if you asked for a review of that removal or of a pause from posting, a record that you asked and how it ended — each kept with your account, seen only by you and, for a review, by a moderator who sees the post’s label and the rule, never who you are, and gone with your account.
• If you asked to join the beta through an invite link, the email you gave and, if you wrote one, your answer to “What do you like to read?” — for the people running the beta to read, until you delete your account or ask us to remove it.
• A record of each AI call made for you — its task, its model, how much it used and its estimated cost — to keep within spending limits. It never holds what was sent.
• Our services’ logs, which record your account’s internal id, your books’ ids and the address of each request, to run and fix the app.

What other readers see

Your books and everything made from them are never shown to anyone else. If two people upload the same book, each gets a fully separate generation — never a shared copy. See the Terms of use for the full rule. What does reach other readers is this, and only this:

• Friends — added only by an invite link you share — see your name, bio, picture and banner; the title of what you’re reading while you share it, never how far; your Four Favourites, recommendation sets and Want to Read as titles and authors; the quotes you make public (a copy of the words, up to 280 characters, with your note and the book’s name — never where in the book); and that you’re reading right now, while you are. When you pass a quote to one friend, that friend alone sees its words, the line you add and the book’s name; when a friend writes back on a quote of yours, you alone see their note — each is kept only until it’s put away, and neither says where you are in the book. Reading a book together shares its title and author, your name, the posts and short quotes you write there, and a line on the pace — never how far you are.
• Every signed-in reader sees what you post in the Agora, but never who you are: each post is labelled only “OP” or “Poster 1”, “Poster 2” within its thread, with its time to the minute and, when you quote your own copy, whether you’d read past the quote. A note you leave for the next reader is shown, with no name, to a stranger who finishes the same book.
• Someone who isn’t your friend yet sees only your name and a monogram.

Who processes it for us

The app sends each service the minimum its job needs — a passage, never a whole book at once — and sends your name or email only to the services that sign you in and deliver its emails:

• Google (the Gemini API) does most of the app’s AI text work. It receives one chapter at a time, to plan a book’s scenes and write its chapter overviews; the book’s title, chapter titles and a short passage from its opening, to set its era and style; the text up to your place, never past it, for a recap; the sentence you’re on with up to two before it, the chapter’s overview and up to three of your own highlights near it, for a question; up to about 150 words of a passage, to direct a picture, and then the picture and that passage, to check it; and the words of an Agora post or a note for the next reader, never who wrote them, to check them against the rules. It also makes a book’s music (Google’s Lyria) from a short description of each track’s mood — never the book’s text — and, when the cover model is switched on, paints a book’s cover from the same kind of written description Cloudflare gets for a picture, never the book’s title or text. We use Google’s paid service: under its terms, Google doesn’t use what we send to improve its products, and it keeps prompts and responses for 55 days to detect abuse.
• Anthropic (Claude) writes Explain’s three doors — “Think about it”, “Explain it back” and “Help me understand”. It receives the passage you chose and its paragraph (for “Help me understand”, only up to the passage’s end, with the chapter’s overview) and, for “Explain it back”, your own explanation. Once for each book you upload, it also receives the book’s title and author and our chapter-by-chapter notes on its characters and places — including a few of the book’s own descriptive phrases, never its text as a whole — to keep the pictures consistent. It never receives your name, email or account id. Anthropic doesn’t train its models on what we send, and deletes it within 30 days — or keeps it for up to 2 years if its safety checks flag it.
• Cloudflare (Workers AI) paints the pictures (and a cover, if Google’s cover model is off or declines). It receives a written description of each picture — the scene, and people by how they look, never by name — and, for a character’s pictures, their earlier portrait. If read-aloud narration is switched on (it’s off today), it would also receive the text of each chapter it reads. Under its terms, Cloudflare doesn’t use what we send to train models.
• The dictionaries — the Free Dictionary API and Wiktionary — receive one word at a time when you look a word up, or when word hints are on, asked by our server and never told who asked; a word’s recording, when you play it, is fetched the same way, from the Free Dictionary API or Google.
• Running the app: Vercel hosts it and keeps its request logs; Supabase holds its database and your files; Upstash keeps short-lived counters for the daily limits, who’s reading right now (for about 90 seconds, and only while you share it), and where the faces are in a picture, as boxes under a scrambled key that names no one (for up to 60 days, so the words can be placed clear of them); Trigger.dev runs the background work that processes a book, given your account’s and the book’s ids; Resend delivers sign-in emails; Sentry receives error reports — an error’s message, where it happened, your browser, and the address of the page or request (which can name a book’s id, or a word you looked up) — to help us fix bugs. Google, and Apple where it’s offered, sign you in when you choose them.

“Device voice”, which reads along as autoplay plays, uses only the voices on your own device: the book’s text isn’t sent anywhere for it.

If you turn on a reminder for your reading time, we keep that device’s push address with the days and time you chose and your device’s time zone, and at that time send it through your browser’s push service (Google, Apple, Mozilla or Microsoft, depending on the browser). The reminder says only “Your reading time” — never a book, a place or anything you read. Turning it off, or signing out on that device, forgets it.

We don’t sell your data, and we don’t use it for advertising. We don’t run ads.

Cookies and your device

We don’t use tracking or advertising cookies. The app sets:

• its sign-in cookie, which keeps you signed in, and, while you sign in, a few short-lived ones that protect the sign-in itself;
• a mark of where you left the app — the book, if you were in one — so it can open where you were (60 days);
• which Agora board you last looked at (a year);
• whether you’ve chosen to open every chapter (a year).

On your device the app also keeps your reading settings; a place not yet saved while you’re offline; the words you looked up and which margin notes you’ve seen, book by book; the pages and pictures of the last few books you read, and any you took with you to read offline; and the app’s own ambient sounds, which are the same for everyone. Everything there that’s your account’s is cleared when you sign in or out, and a book’s own when you delete it.

Deleting your data

Deleting a book removes it, its text, everything generated from it, your place in it and your highlights and notes from it — on our side, and on the device you delete it from.

Settings, on your profile, has “Delete account…”. It removes your books and everything generated from them, your highlights and notes, your profile, your friendships and everything else listed above, and can’t be undone. Your Agora posts are taken back first — their words are never shown again — and your threads stay, with no name on them, so other readers’ replies in them aren’t lost; the board deletes them in time like any other thread. Files a background job was still writing when you deleted a book or your account are cleared by a daily sweep. What a service above keeps for its own set period is deleted on its schedule.

Children

The app isn’t directed at, or knowingly used to collect information from, children under 13. The Agora is for readers 18 and over: when you confirm you are, the app keeps when you did, on your own profile, and never shows it to anyone.

Changes to this policy

We’ll update the date above whenever this changes.

Contact

Questions about this policy, or a request about your data: admin@conjure-reader.app.