SigilAtlas

The apps

The studio and the field.

Two forms of one instrument: the desktop, where an archive is built and worked — free while in beta; the web, where a world is visited.

Both apps are in beta · the network is by invitation — where things stand

The studio — desktop

Your archive lives on your machine and stays there. The desktop app reads photographs locally; bringing them in uploads nothing, anywhere.

Privacy here is the simple kind: your photographs never leave your machine unless you publish. The models that read images, faces, and words all run locally — which is also the hardware bar. SigilAtlas is a Mac app and wants Apple silicon, M1 or newer; with several models running side by side, anything less will crawl, if it runs at all.

The SigilAtlas desktop app in Remember mode: the archive laid into horizontal timeline strips, hue contrasts listed on the right, saved narratives on the left
The studio in Remember mode: the archive laid into strips, hue contrasts on the right, saved narratives on the left rail.
Download the desktop app

macOS · Apple silicon (M1 or newer) · beta

For photographers: the Lightroom studio

Your archive is not a folder tree; it is a body of work, and this is the instrument that shows its structure. Lightroom stores and edits; SigilAtlas reveals, sequences, and re-enters. The studio reads a Lightroom Classic catalog directly — bring in an archive without exporting anything. Sync on demand from SigilAtlas whenever you want to bring in Lightroom’s latest edits, and the map repaints in place. Gather a collection in SigilAtlas and it writes back into Lightroom as a real collection, keywords and all, ready for Lightroom’s export. SigilAtlas never touches your originals. The plug-in is a professional workflow tool — a subscription, purchased inside the app.

The resident guide

An agent lives inside the studio — not a chatbot beside it. It works the instrument through the same controls you do: speaks phrases into the field, places subjects, opens narratives, looks at what it has made and adjusts. It teaches by showing, not telling — watch it move through your archive, and after a few passes the pattern becomes yours. You bring your own model account (Anthropic or OpenAI). The guide is included free for the first hundred atlases; after that it becomes a one-time purchase you own.

The Dialogue pane: a conversation with the resident guide beside a live canvas it is reading — the guide describes what it likes about the archive and why the canvas reads cleaner than the map
The guide in dialogue over a live canvas, reading the field it has made. The pane's last line names your model account and the running cost.

The field — web

The public, immediate form — an archive published as a world anyone can walk.

Privacy here is tiered by who you are to an archive. Publishing is an explicit act — and an explicit contribution: your photographs join the field as building blocks, raising the resolution and the diversity of the patterns it can form. From there, what a visitor sees is a matter of relationship:

anyone
Walks the whole field, but sees each photograph as a 32-pixel square crop. Enough to tell it’s a photograph, even that it’s a face; not enough to look closely. And there is no way to zoom further.
the invited
See an archive at the resolution its owner chose when publishing — say, 256 pixels across: legible without weighing the field down. A light panel opens the actual photograph, uncropped, not the square tile.
you
See your own archive at the resolution you uploaded. In the high-resolution view, your world is exactly the archives you’ve been invited into — no further.

Between worlds, photographs are doors. To visit your mother’s archive you type @mama — the name you gave her. Her cell gathers; inside it, one photograph is highlighted: a door. Click it and the field crosses into her world; the same door leads back to yours. Doors exist only where you’ve been invited.

The SigilAtlas web app: the live field of photographs with the Narrate panel docked on the right
The field: the live archive with the Narrate panel — spells, contrasts, brightness — docked on the right.
Open the web app

Coming

The commons. Publish galleries into a shared field — a superposition of everyone’s contributed photographs, each a tiny tile. Your world, with galleries within, that anyone can visit and the invited can enter. Not ready yet; coming.

Shared sessions. Invite someone into a slice of your atlas without publishing anything: your machine hosts the session, they visit through a private link, and you talk over the pictures. In the works.

Specialist themes. Tell the instrument who you are, and its labels rename into your discipline’s terms — a physicist’s chrome, a linguist’s, a semiotician’s — each term glossed back to plain, the spell language unchanged underneath. The resident guide reads the same declaration and speaks your vocabulary.


Where things stand

Both apps are in beta, and the network is by invitation. That is the design, not a launch tactic: the people who can look closely at a archive are the people who appear in it.

What costs what, plainly: the desktop app is free while in beta. The resident guide is included for the first hundred atlases, then a one-time purchase. The Lightroom plug-in is a subscription from the start — it is a professional workflow tool. Whatever is sold one-time is yours: no feature you bought outright ever starts renting itself back to you.

An invitation runs like this:

  1. You invite someone you have photographs of. They receive a link.
  2. They register through that link — there is no public sign-up.
  3. They download the desktop app — Mac, Apple silicon — and sign in.
  4. They bring in an archive of their own, from a folder or from Apple Photos.
  5. They find you in their photographs and give you your name.
  6. They publish. Your face in their archive is now a door: they step through into your world.

Written out, it reads like a protocol; walked, it takes minutes. And every step is there for a reason: to get into your world, they need to have theirs, where you are known and named. A door only ever opens between two points of view.

So this is a network — but one with no textual layer. No likes, no posts, no comments; for now, no narrative layer either. What moves through it is point of view: entering someone’s archive, you see what they noticed — nothing more, nothing less.

Who makes it

SigilAtlas is a startup at its smallest: one person, building with Claude and, at times, Codex. A new kind of company, where the human contribution is the idea — the point of view — and most of the code is written in conversation with non-human points of view.

Why it exists →