Every post you saved, in a library you can search.

No, and we could not accept it if you offered. Connecting runs X's own authorization flow — OAuth 2.0 with PKCE — so you approve the request on x.com and X hands us a token afterwards. Your password never reaches this app.

The token is limited to four read scopes and nothing else: tweet.read, users.read, bookmark.read, offline.access. It is encrypted before it is stored, and deleting your account revokes it at X rather than merely forgetting it here.

No. We never request a write scope, so the token we hold is physically incapable of posting, replying, liking, following, unfollowing or messaging — there is nothing to misuse and nothing for us to promise not to do. It can read your posts, your public profile and your bookmarks. That is the whole surface.

Yes, progressively. Pro imports your newest 500 the moment you connect, then reaches back another 250 every week — so a 10,000-bookmark archive fills in over about 38 weeks, with a progress bar in the app the whole time. Backfill keeps reaching back until it has 20,000 of your history.

It is deliberate rather than a limitation we are embarrassed about: importing everything on day one would cost us tens of dollars before your first payment, which is exactly the shape of business that has to raise prices later. New bookmarks you save from now on arrive the next day regardless — backfill only affects how fast the old ones appear.

It is stored with its author, its canonical x.com link and the date you saved it, then a model gives it one of 12 fixed categories, a few tags and a one-line summary. Those three things are ours, not X's, so they are shown in their own labelled row and never folded into the post text. After that it is searchable, filterable by category and tag, and it appears in your next digest.

Because reading bookmarks is metered. X bills us $0.005 per post returned, so your 100 bookmarks are $0.50 of post reads plus a lookup for every author we have not seen before — roughly $0.55 per free signup, paid by us, before you have paid anything.

100 is a number we can afford for everyone who wants to try this. Uncapped, a single signup with a 5,000-bookmark history costs about $26 on the spot, and a free tier shaped like that fails on its best day. We would rather publish the limit than discover it in public.

$6 a month, or $60 paid once a year, which works out to two months free. It buys the newest 500 bookmarks up front, another 250 of your archive every week, a daily digest instead of a weekly one, and CSV export.

No. The price is flat, and what we spend at X is our problem to manage, not a line item we pass to you. The fair-use threshold is how we bound it: 1,000 newly saved bookmarks a month on Pro, 200 on Free. Crossing it starts a conversation about what you are doing — not a surcharge and not a silent slowdown.

Because checking for new bookmarks costs money every time. We ask X for your single most recent bookmark and compare it to what we already have — one post read, $0.005. Pro does that daily; Free does it every 7 days, which is what keeps a quiet free account down to a couple of cents a month. The digest cadence follows the check, because emailing you daily about a library we only look at weekly would be theatre.

It goes from here within 24 hours. Every night we ask X's batch compliance endpoint which of the posts we store have been deleted, withheld or suspended, and act on the answer before the next digest goes out — so a deleted post cannot turn up in your inbox afterwards.

What you see is a tombstone rather than a hole: the row stays in your library marked as no longer available on X, with the content removed. Silently dropping a bookmark you remember saving is worse than telling you it is gone. This is X's requirement for anyone storing post content offline, and it is also the behaviour we would want.

Two services, both named because you are entitled to know. Anthropic receives batches of post text to produce the categories, tags and summaries — without your handle, your email or your user id attached, because classification does not need to know whose library it is. Our email relay receives the rendered digest, which is what sending it to you requires.

That is the entire list. Your bookmarks are not sold, not used to train anything of ours, not enriched with third-party data and not shown to other users. The privacy policy spells out both processors and what each one receives.

No, and we work to make that obvious. X's display rules forbid altering post content, and a generated label sitting inside a post body reads as exactly that. So categories, tags and summaries always render in their own row, in their own colour, with their own label — in the app and in the email. Even the CSV export prefixes those columns with ai_. The post text itself is never edited, trimmed or rewritten.

No. X's policy forbids storing or aggregating location data separately from the post it belongs to, so we do not extract it at all — there are no geo columns in the database to fill in. There is no map view, no "bookmarks near you", and no plan for either.

One button in Settings: we revoke our access token at X, then purge your bookmark list, the categories, tags and summaries generated for you, your sync state, your digest history and your stored tokens. No support ticket, no cooling-off period, no "deactivated" state that quietly keeps everything.

Two things survive, and we would rather say so than claim a clean sweep. Post text itself sits in a store shared by everyone who bookmarked the same post, so deleting your account removes the link between you and those posts rather than other people's copies. And the log of what we spent at X stays as a financial record, with your user id stripped out of it. The privacy policy sets out both.

Yes — 500 posts per day, on Pro. That number is X's, not ours: their developer policy permits handing up to 500 post objects to each person who uses a service per day, by non-automated means. So export is a button you press, the count is enforced on our server, and a large library comes out over consecutive days — 2,000 bookmarks in 4 days, for instance.

The file is CSV, one row per bookmark: post id, canonical link, author handle, when it was posted, when you saved it, the post text, and the category, tags and summary in their own ai_ columns. Posts we have tombstoned are left out, because content deleted on X may not be redistributed.

Not today. Pushing post content into a third-party service on a schedule is precisely what X's redistribution rules restrict; the exception they spell out is a download the person asks for. So we do the download well and leave the last step to you — the CSV imports directly into Notion or a spreadsheet, and our free in-browser converter turns it into Markdown for a notes app. If that changes, it will be because we asked X and got a written answer, not because we decided the rule probably did not apply.

Price only. Monthly is $6, which is $72 across a year. Annual is $60 billed once — $12 less, or two months free, depending on which way you like to read it. The caps, the digest and the export allowance are identical.

No. The free plan needs your X connection and nothing else — no card, no expiry date, no countdown. It imports your newest 100 bookmarks, sorts them and sends a weekly digest for as long as you want it.

You keep the library you already have. Moving to Free reapplies the free caps — a weekly digest instead of a daily one, a check for new bookmarks every 7 days, no CSV export and no further backfill — but nothing already imported is deleted for changing plan. The only things that remove a bookmark are deleting your account, or the post disappearing from X. Self-serve checkout is not switched on for this deployment yet, so there is no subscription to change here today.