Legal

Terms of service

What Sift commits to — read-only access, published caps the code actually enforces, and your data leaving as easily as it arrived — and, just as plainly, what it does not promise yet.

Drafted 9 September 2026Not yet in force

This is a draft for the operator to review with a lawyer.

It was written by the people who built the software, from the code and from our reading of X's Developer Policy, so the descriptions of what is stored, sent and deleted are accurate. It has had no legal review, it names no legal entity, and it is not in force. Read it as a truthful description of how the system behaves — not as a contract, and not as advice about your rights.

1What this document is

These are the terms on which Sift intends to be offered: what the service does, what it commits to, what it asks of you, and where its limits are. It is written to be read, so it says "we" and "you" and avoids clauses that exist only to look like clauses.

It is also unfinished. The parts that describe the software are accurate. The parts a contract normally leans on — liability, warranties, governing law, refunds — are not drafted, on purpose, and are listed at the end for the operator's lawyer to write.

2What the service does

You connect your X account. Sift reads the posts you have bookmarked, stores them, gives each one a category, a few tags and a one-line summary, and makes the result searchable. It emails you a digest of what you have saved — daily or weekly, depending on your plan — and Pro can export the library as a CSV file.

What it is not: a backup of your X account, a scraper, a way to reach content you cannot already see, or a substitute for X. It reads what X's API returns for your own bookmarks, and nothing else.

3The X connection

  • Read-only, by construction. We request tweet.read, users.read, bookmark.read and offline.access. There is no write scope on the connection, so nothing we hold can post, reply, like, follow, unfollow or message as you — whatever our code might one day ask it to do.
  • Yours to withdraw. You can revoke our access from your X account settings at any moment, and deleting your Sift account revokes it for you and purges your data.
  • Your X account remains yours to keep in good standing. X's own terms still apply to you and to the content you saved. Using Sift does not change them, and does not give you rights over other people's posts that you did not already have.
  • An independent product. Sift is not affiliated with, sponsored by or endorsed by X Corp.

4Plans, and the caps we commit to

These limits are not a marketing summary of the product; they are the numbers the software enforces, read from the same configuration the sync engine and the plan gate use. If a cap changes, this table changes with it.

FreePro
PriceFree$6 / month or $60 / year
Import100 newest, once500 newest, then +250 a week to 20,000
DigestWeeklyDaily
Check for new bookmarksEvery 7 daysDaily
Bookmarks stored, lifetime2,00025,000
Fair use, new saves per month2001,000
CSV exportNot included500 posts per day

Scroll sideways to see the whole table.

The reason the free import stops at 100 is cost: X bills per post returned, so importing a library is a real expense we pay before you pay anything. The reasoning is set out in full in the pricing section.

5Fair use

Every bookmark we read costs us money, so both plans carry a monthly threshold on newly saved bookmarks: 200 on Free, 1,000 on Pro. It is a soft line. Crossing it means we will get in touch about slowing your sync or moving you to something that fits — not a surprise charge, and not a silent degradation you have to discover.

There is also a hard ceiling underneath it: a per-account monthly budget for X API spend, and a global daily kill switch. These exist so that a bug in our own cost control cannot run up an unbounded bill, and in normal use you will never meet either.

6Prices and payment

  • Pro is $6 per month, or $60 billed once a year — $12 less than paying monthly.
  • Prices are in US dollars and exclude any sales tax, VAT or GST that applies where you are.
  • There is no setup fee, no per-seat pricing and no usage surcharge. What we spend at X is ours to manage.
  • Moving back to Free reapplies the free caps — weekly digest, no CSV export, no further backfill — and does not delete the bookmarks you have already imported. Only deleting your account, or a post disappearing from X, removes content.
  • Not yet drafted: renewal and cancellation mechanics, refunds, any trial, invoicing, and the statutory withdrawal rights that apply to consumers in some markets. Those are open items below rather than promises made here.

7Content, and who owns what

  • The posts are not ours and not yours. They belong to the people who wrote them and remain subject to X's terms. We store copies so you can search what you saved, and every one links back to the original on x.com.
  • The categories, tags and summaries are ours, generated from post text. You may use and export them freely as part of your own library. They are always shown separately from post content, because presenting generated text as part of a post would be altering that post.
  • Your bookmark list is yours. We claim no ownership of it, do not sell it, and do not use it to train models of our own.
  • Generated text can be wrong. A category or summary is a model's guess. Treat it as a filing aid, never as a statement about what a post says — the post itself is one click away, which is the point of the link.

8What you may not do with it

Most of this section exists because X's Developer Policy binds us, and we can only keep those commitments if the service is not used to route around them.

  • Do not redistribute exported post content onward — republishing it as a dataset, an API, or a public archive. The export exists so you can use your own library in your own tools.
  • Do not work around the 500-posts-per-day export limit, with multiple accounts or by automating the download. That cap is X's and we have to honour it.
  • Do not use the service to build a competing dataset of X content, or resell access to what it stores.
  • Do not attempt to reach another user's library, probe the app for vulnerabilities without asking first, or automate against it in a way that degrades it for others.
  • Do not connect an account you are not entitled to use. One person, their own bookmarks.

9Availability, and what is outside our control

Sift depends on X's API, and X states plainly that its pricing, rate limits and policy are subject to change. If X changes what it charges or what it permits, this product has to change too — including, in the worst case, a cap moving or a feature going away. We would rather tell you that now than describe an entitlement we cannot guarantee.

There is no uptime commitment yet. Syncing runs once a day per account and a missed run is picked up by the next one; a digest with nothing new in it is never sent at all. If a published cap changes, the change is published — the same numbers appear on the pricing section, in this document and in the code, so a quiet reduction is not something we can do without it being visible.

10Ending it

  • You, at any time. Delete your account from Settings. We revoke our token at X and purge your data, as described in the privacy policy.
  • Us, if we have to. We may suspend an account that breaches the acceptable use section above, that is using the service in a way we are contractually unable to support, or where payment fails and is not resolved. Where the reason is not a security matter, we will say what it is and give you a chance to export first.
  • Either of us, if X withdraws access. Our exposure to a policy breach is not a fine, it is the API being switched off. If that happens the service cannot function, and we will say so rather than let it fail quietly.

11What still needs a human

Everything above describes what the software actually does, which is the part we can state accurately. The following are legal judgements, not engineering facts, and a qualified adviser has to make them before this page can bind anyone.

  • The operator's legal entity, jurisdiction and governing law, plus how notices are served in each direction. Every clause below depends on this being settled first.
  • Warranty disclaimer and limitation of liability. Deliberately absent. These are the two clauses most often copied verbatim and most often unenforceable as copied, and they interact with consumer-protection law in ways that vary by market.
  • Refunds, cancellation and any trial. A proposal for the operator to accept or reject, not a commitment made here: a 14-day money-back window on the first Pro payment. It is simple to honour, and the free plan already lets someone try the product with 100 of their own bookmarks. Nothing on this site promises it today, and nothing should until the operator decides.
  • Subscription mechanics. Renewal dates, auto-renewal disclosures, price-change notice periods, proration on upgrade, dunning on a failed payment, and consumer withdrawal rights where they apply. Checkout is not live yet, so none of this is settled.
  • Tax treatment and invoicing, including where the operator is registered and which customers are charged tax.
  • A copyright and takedown route. We store copies of other people's posts; someone will eventually ask us to remove one. The nightly deletion sync handles content removed from X, but a direct complaint needs a named recipient and a documented process.
  • Any service-level commitment, if the operator wants one — uptime, digest delivery windows, and what happens when they are missed.
  • Minimum age, and assignment on a change of control. Both standard, both dependent on the jurisdiction chosen above.
  • Dispute resolution. Whether disputes go to a named court, to arbitration, or through an informal step first — and whether that is even permitted for consumers in the operator's markets.