Privacy
Your privacy.
What the Aggregator keeps, who else handles it, and what it never does.
What we keep, and why
Your sign-in. Your email and a one-way scramble of your password. We never see or save the password itself.
It lets you sign in.
Your topics. Each one written in your own words, and the order you put them in.
They decide what is in your brief and in what order.
Your sources. The publishers you added, whether each is free, behind a paywall or one you subscribe to (a choice, never a login), whether you want all of a site's stories shown, which sites you pinned in your brief, and their order.
They tell us whose feeds to read for you, which sites to treat as paywalled, which stories to show whatever your topics, and which site links to keep one tap away.
What you like to be called. A name, if you give one.
It is optional.
A record of each Claude call. The date, which job it did, the model, how much text it used, what it cost, and whose request caused it. Never the text sent or received.
It keeps spending under a daily limit.
A count of each brief you open. How many stories it held and how many were not checked. Never which stories.
It shows whether checking keeps up with the daily limit and what it costs.
Which of your topics each story falls under. For each story in your brief, which of your topics it matches, or none, whether it hit anything you would rather skip, and when we last tried. Never the story itself.
So your brief can group stories by your topics without asking Claude again each time you open it.
Things you'd rather skip. Each one written in your own words as you confirmed it, and its order.
So we leave those stories out of your brief and count them.
Your What I see choices. Your paywall level, whether the major story heads-up is on, whether your brief shows tiles or a list, how it is arranged (by topic, by source or latest), and whether you have seen the pointer to What I see.
So your brief follows the choices you made.
Stories that got past something you said to skip. When you tap Not for me on a story that one of your skip items should already have caught: which story it was and which skip item. Never the story or your words.
So we can see where leaving stories out falls short.
Stories you saved. Which stories you chose to keep: each story's link and when you saved it. The headline and summary on your Saved list are read from the shared copy; we never keep a copy of the article.
So your Saved list is there on every device, only for you. It is never used for suggestions or ranking, and never shared.
Stories we showed you in the last 7 days. For each story we put in your brief: its link, when you last saw it, and a short code that changes when its headline or summary changes. Never the headline, the summary or the article.
So we can tag a story Updated when it changed since you saw it, and tell you when you are caught up. It is never used for suggestions or ranking, never shared, and each one is deleted 7 days after you last saw it.
Stories you added or hid. Which stories you chose to add to your brief or hide from it: each story's link, whether you added or hid it, and when. We never keep a copy of the article.
So your choices hold from one visit to the next, only for you. They are kept for 7 days at most and never shared. We use them only to suggest a change to your topics, which you can accept, edit or ignore, or turn off in What I see, and never to rank your stories.
Stories you opened. Which stories you tapped to read: each story's link and when you opened it. We never keep a copy of the article.
So a story you already opened can say so when you come back. They are kept for 7 days at most and never shared. We use them only to suggest a change to your topics, which you can accept, edit or ignore, or turn off in What I see, and never to rank your stories.
Topic suggestions. When a suggestion to sharpen a topic shows in What I see, we send Claude the topic and up to 6 headlines of stories you hid under it, once, to word the change. Never your name, email or the articles. Whether you turned suggestions off is kept with your settings.
So Claude can put the change in plain words for you to accept, edit or ignore. What we send is never stored, and never used to rank your stories.
Sites you tried and dropped. When you try a site, we keep it in your sources, marked as trying. If you drop it, we keep its name in a short list only you can see, so you can bring it back. Nothing else about it is kept.
So you can change your mind. We never share your sources.
Topics you folded. Which topics you folded, kept on this device only.
So your brief opens the way you left it on this device. We never receive it.
Sites we told you about. Which sites we told you could not be read, or can be read again, and whether you closed the notice, kept on this device only.
So we say it once and don't repeat a notice you already closed. We never receive it.
Today's mood and source. Today's mood and the sites you narrowed your brief to, kept in this browser tab only.
So going back from a story keeps them. They are forgotten when you close the tab, and we never receive them.
Searching for sources. The words you type in Find sources, or the name of a site you add that we cannot read on our own, go to Anthropic, who run the web search for us. Nothing you typed is kept except that a search was made and what it cost. Suggested sources are not stored.
So we can suggest written sources for a subject, and find the page where a publisher lists its own feeds.
Pictures from publishers. When a story's own feed carries a picture, your phone loads it straight from the publisher's site. That publisher can see your device's address and when the picture was asked for, whether or not you open the story. We send them nothing else: not our address, not your name or email. We never keep a copy of a picture, only the address the feed gave.
So a tile or a list entry can show the picture the publisher chose for its own story. A story with no picture, or one that fails to load, tells its publisher nothing.
Shared with every reader, not tied to you. Each publisher's name and feed addresses (a main feed and sometimes section feeds), and each story's headline, short feed summary, link, picture address and check result. A story is stored once for everyone. Full articles are never stored.
So each story is read and checked once, however many readers follow it.
Shared with every reader, not tied to you: sites we could not read. When a publisher's site turns our reader away, asks apps like ours not to read it, cannot be reached or says to try again later: the reason, the date we first saw it, and, for a site that turned us away, how many times we asked. We look again once a day, or once a week for a site that turned us away. Never who asked.
So your brief can say plainly which sources could not be read and when we will look again, and so we know which publishers to ask for permission.
Counted for everyone, not tied to you. Each day, how many times we sorted a story into topics and how many of those tries failed. Two numbers for the whole day. Never a reader, a story or any text.
So we notice when sorting keeps failing and fix it.
Who else handles your information
Supabase. Keeps our database and signs you in.
Your email, the scrambled password and everything in the list above.
Resend. Sends the invite and password-reset emails.
Your email address and the email's text.
Vercel. Hosts the site.
Each page request, as any host does.
Anthropic. Runs the story check, matches stories to your topics, understands words you type and runs the web search for Find sources.
Story text (the headline and summary, or the full article where the publisher allows it), your topic names and words you type, such as an exclusion or what you search for in Find sources, and the name of a site you add. It does not get your name or email address.
What we never do
No ads. No ad tracking. Your settings and saved stories are stored only for you.
- We never store your password or any publisher login.
- We never store full articles.
- We never send push alerts.
- We never sell your information.
Deleting your information
To delete your account and everything stored for you, ask the person who invited you. Your topics, sources and settings go with it. The Claude usage record stays, with the link to you removed, so cost history stays complete.