Straight answers, including the awkward ones.

It means Sally can keep a detailed record of a file without keeping the file. A library catalog card describes a book down to its page count without the library owning your copy — same idea. When you run the Sally app on your own computer, your photos stay exactly where they are, and what travels to Sally is the description: the fingerprint, the size, the dates, the folder it lives in, and for photos the tiny preview your camera already tucked inside the JPEG.

Only the ones you deliberately drag into the browser. Those are stored with Sally and can be downloaded again from the item’s page whenever you like. Everything the local app finds is cataloged in place and never uploaded. You never have to guess which is which: every item’s page says either “Stored with Sally” or “Cataloged in place”, and lists every location that file is known to exist in.

Never. Sally reads; she does not rearrange. The local app opens files to fingerprint them and writes nothing back into them. Removing an item from your library removes Sally’s record of it — the file on your drive is not touched.

Nothing until you say so, and even then nothing is deleted from your disk. Sally groups copies by what they look like rather than what they are named, so resized and re-saved versions are caught too. She suggests the best copy — sharpest, largest, actually in hand — and merging a group leaves you with one tidy item that still remembers every place that photo exists. If two pictures genuinely are different, tell her so and she stops asking.

The catalog stays. You keep the fingerprints, dates, labels and the record of which drive that file was on — which is exactly what you need in order to know what you lost, or to find it again on the copy in the closet. Worth saying plainly: Sally is a catalog, not a backup. She will happily show you that your only copy of something lives on one aging drive, which is an uncomfortable and extremely useful thing to learn.

The website accepts files up to 50 MB each, which covers photos and documents comfortably. Bigger things — mostly video — are the local app’s department: it catalogs them where they sit, at any size, without uploading a byte of the file itself.

Yes. Open sallydoesdata.com in your phone’s browser, sign in, and tap the add box — your normal photo picker opens, and photos upload with their dates and locations intact. Reviewing, searching, collections and sharing all work on a phone screen too, which is genuinely the best way to answer a few questions while waiting for a coffee.

No, and not for a while. The website works in a phone browser, and one small operation can only look after so many things at once. The sally command-line app is desktop-only, because that is where the hard drives are.

No, and please do not. Ask Google for a Takeout export of Google Photos, unzip it, and run sally takeout on the folder. Sally reads Google’s sidecar files, so your photos arrive with their real taken-date, GPS, your own titles and descriptions, album names, and the people you tagged — the metadata that usually gets dropped on the floor in transit. Your Google account carries on exactly as before. This is an index, not an eviction notice.

iCloud Drive is one of the places sally go looks, so anything synced there is cataloged like any other folder. The Apple Photos library itself is a sealed bundle that Sally will not root around inside, so if your pictures live in Photos, export the ones you care about first — File, then Export, then Export Unmodified Original, which keeps the dates and GPS — and point Sally at that folder. One warning for laptops running Optimize Storage: a photo that exists only in the cloud has nothing local to fingerprint, so download the originals before scanning.

Request your information from Meta in JSON format, unzip it, and run sally meta on the folder. Post dates, captions and album names come along, and Sally quietly repairs Meta’s mangled accents on the way in, so “Sofía” goes back to being Sofía.

Not yet. The Sources page lists those with a “Coming soon” chip rather than a hopeful checkbox. Today there are three ways in: the website, your phone’s browser, and the local app — which does read your Dropbox, iCloud Drive, OneDrive and Google Drive sync folders, since on your computer those are simply folders.

You. Everything is private by default and stays that way until you deliberately share something. There is no feed, no public profile, and no browsing anybody else’s library. Two privacy levels named Community and Open data appear in the app marked “Future”; they are not switched on and they do nothing today.

You make a collection — Sally’s word for an album — and click “Share by link”. She stops first and tells you in plain words exactly what is about to happen, then gives you a web address to paste into an email or a text message. Whoever opens it needs no account and nothing installed, and sees only that collection: its name, its pictures, their dates, and nothing else of yours. Search engines are told to stay out.

It is unlisted rather than password-protected, like an unlisted phone number. Anyone holding the link can open the page, including whoever it gets forwarded to, so share accordingly. You can switch the link off at any moment, and setting an item back to Private pulls it off every shared page instantly, even if the link is still live.

No, and no. Sally does not sell personal information, does not hand your content to model trainers, and runs no advertising trackers on the product. The business model is deliberately boring: people pay for the shelf, not with the stuff on it.

The privacy policy says the same thing in the language lawyers prefer.

Sally speaks MCP, which is the standard way to give an assistant like Claude access to something. You create an access token in Settings, paste it into your assistant’s connector settings, and then ask real questions: find my photos from the summer of 2019 with Mom in them, or which of my documents mention the house. The connection is read-only — your assistant can search your catalog and read what you wrote, but it cannot change or delete anything, cannot download original files, and stops the second you revoke the token.

Because an assistant can only work with what it can read. “IMG_2847.jpg, 3.2 MB” tells it nothing worth knowing. “Fireworks with Mom, July 4 2019, Logan Square” tells it everything. The Fidelity score is really a progress bar toward the day your own tools understand your own life — which is the entire point of the exercise.

Nothing today — no plan is being charged yet. When billing starts, the free plan keeps 5,000 items, two connected sources, every export format and read-only AI access. Plus will be $6 a month billed yearly, Pro $24 for people who catalog on behalf of others, and there are 500 one-time Founding Archivist seats at $199 for Plus for life. No card is needed to start, and prices are published before anyone is charged.

The whole grid, including the rows where every plan says yes, is on the pricing page.

No. It is a real plan with a real ceiling: 5,000 unique items, two sources, and every export. Duplicates count once, which is a friendlier way to count than most. If your life turns out bigger than that — most lives are — you will hear the price from the pricing page long before you hear it from a checkout.

Not yet. Signup asks you to confirm you live in the US, because a one-person operation should not pretend it has GDPR compliance, EU data residency and international support figured out on day one. It is a limit, not a verdict on your country, and it is meant to be temporary.

If you are elsewhere and want in, write to hello@sallydoesdata.com and say where you are. The list of who is waiting is how this gets prioritized.

One person, one small server, and an unreasonable feeling about lost family photos. That is exactly why this page is so blunt about what is missing: it is easier to say what is not built than to be found out later. It is also why every export format exists. A small operation should be easy to leave.

That is most of the point. A labeled, dated, deduplicated catalog is the difference between an heirloom and a hard drive nobody can face. Today the way to hand it on is an export: the CSV opens in any spreadsheet, the JSON-LD uses vocabulary other systems already understand, and the SQLite file will still open in twenty years with no help from us. Put one next to your will, along with where the drives are and how to get into them. A proper legacy-contact feature is not built yet, and saying so is better than implying otherwise.

The free digital estate checklist is the interim version, and it needs no account.

You would get at least 90 days of notice, exports would keep working the whole time, and a Founding seat inside its first year would be refunded pro rata. That is a cheap promise to make and an expensive one to break, which is precisely why it belongs on a public page rather than in a support reply.

The short, honest list, so nothing surprises you later.

• No email of any kind yet: no confirmation, no password reset.
• Video is cataloged and labeled, but not previewed.
• The website takes 50 MB per file; the local app has no such limit.
• Dropbox, Google Drive and S3 connections are listed as coming soon, and they are.
• The Community and Open data privacy levels are marked Future and do nothing.
• The one-click bundle of files and labels together is still being built.
• United States only, ages 13 and up.