Privacy

You are about to point this at years of work, some of it a client’s. Here is exactly what happens to it, in plain terms, with nothing left out.

What leaves your computer

Six numbers, and nothing else. No account, no licence check, no update ping.

Exactly one part of Yellide is able to send anything at all: the counter described below, which you can stop at any time. If any other part were ever made to send something, an automatic check would refuse to let that version out.

The one time anything leaves your machine is when you ask Claude to look at your pictures. Those images go to your own AI provider, on your own subscription, in your own conversation. Yellide is not in the middle of it. Never ask, and nothing ever leaves.

If you never ask for descriptions, nothing ever leaves.

What we collect

No personal data at all. No analytics on this website either: no tracking pixel, no cookie.

One thing is collected, described in full below: six numbers that add your totals to a figure on the front page. Nothing in them is about you, which is why there is no consent question. You can still stop it.

The counter, and why it does not ask permission

We want a figure on the front page showing what Yellide achieves across everyone using it: files indexed, by kind, and how far content search has improved. Cloudflare already counts visits to this website. Nothing counts whether the tool then worked.

Everything it sends

Six numbers, the version, and one random label. No filename, no folder, no photograph, no search term. No address, no computer name, no operating system, no country, no time zone. Anything else is refused outright rather than quietly ignored, so a later version cannot start sending more without this page changing first.

The random label is not stored either. It is scrambled on arrival with a key held on the server, in a way that cannot be undone, and dates are kept only as a month. What ends up in the database is a string nobody can trace back and six numbers. Once a day at most, in the background, only while you are using Yellide anyway.

Why there is no consent question

India’s Digital Personal Data Protection Act governs personal data: data about an identifiable individual. Consent is what makes processing that data lawful, and it has to be asked for properly.

None of the six numbers is about a person. There is no name, no email, no phone, no address, no device, no location, and nothing that could be matched against any of them. The one identifier involved is random, made on your machine, and is destroyed on arrival. Nobody holding this database, including us, can work out whose row is whose or that any particular person has Yellide at all.

So there is nothing here to consent to, and we do not pretend otherwise. A consent box over data that is not personal is theatre, and it teaches people to click past the ones that matter.

You can still stop it

Not because the law requires it. Because it is your machine.

Yellide tells you about the counter once, after your first index, in plain words. Tell Claude stop the Yellide counter at any time, before or after that, and it stops and your row is deleted. No account, no email, no form. Say it again to turn it back on.

All of it is public: what is stored, and what receives it. If any of it looks wrong, say so.

What gets written to your drives

Two things, and only two.

The index itself

A single SQLite file in the normal place for application data on your system:

Delete that file and Yellide has forgotten everything. Nothing else is left behind.

A marker on each drive you index

When Yellide indexes an external drive it writes one small file to it:

<drive>/.yellide/volume-id

One random identifier, nothing else. It is what lets Yellide still recognise a drive after you rename it, so it can say “that clip is on KERALA_2023, and that drive is not plugged in.”

You can turn it off. In Claude Desktop: Settings → Extensions → Yellide → Configure, then Never write anything to my drives. In any other MCP client, set YELLIDE_NO_DRIVE_MARKER=1 in the server’s environment.

Indexing and searching are unaffected. You lose only the ability to recognise a drive you have renamed.

What Yellide never does to your files

Why deleting your work is not something it can do

Automated tools have destroyed people’s archives. A sentence on a page promising otherwise is worth nothing, so here is the whole list instead. These are every write and every delete the software performs. There are no others:

Every one of those deletions is of a file Yellide made moments earlier, in the system scratch folder, under a name only that run could have chosen. None of them can name a file of yours.

Yellide runs three programs that already come with macOS: sips and qlmanage to make thumbnails, and open to show you a file in Finder. Each is handed your filename as its own separate piece of information, rather than pasted into a line of text that the Mac then reads back and obeys. So a file named ; rm -rf ~ is only a file with an odd name. Yellide never builds an instruction out of your filenames, and never treats text as something to run.

None of that is a promise you have to take. An automatic check holds it: every write listed above is named there, and a version that writes anywhere else cannot be released.

Private documents

Anything you mark private is recorded by type only, never by its contents, and reduced to a bare label on export. Marking one copy marks every identical copy.

Why you can check all of this

The source is MIT-licensed and public. That is the only real proof any of the above is true, which is why it is the licence.

github.com/urbanmorph/yellide