The Space Studio is where you build. A surface with depth, connectors that carry meaning, and a whole document behind every node.
On a plane, an idea has a position and it has neighbours, which is the entire vocabulary available. Everything else you know about it has to be recorded somewhere else and carried in your head.
Add a third axis and those things become properties of the picture itself, where distance and direction both carry meaning. A grouping holds together when you move around it, because it is a grouping in space rather than an accident of layout.
This is not a failing of flat diagrams. They are precise, portable and often exactly right, but they run out of room to say things, and most people have been compensating for that quietly for years without ever calling it a problem.
Depth is the part of this category nobody else has needed, and there is a good reason for that. A graph you only look at does not require it, whereas a graph you're working inside does.
Stopped, and with the panels open, it looks like this.
There is no upload to regret later and no proprietary file to be trapped in. The structure is a folder of Markdown and conventional files that happens to know how it is arranged.
Start empty, load a Template, or ask an AI to draft the whole thing. Our templates are a vocabulary rather than content: node types and connector types, ready for you to fill. They exist to remove the blank canvas rather than to hand you somebody else's structure.
That is a node. Type its title and it exists, as a real file in your folder from that moment on. Every node is a full document underneath: notes, attachments of any kind, properties and a classification you defined, rather than a label with a tooltip behind it.
Put [[a node name]] in any note and it becomes a working link to that node's page. It is a hyperlink rather than a connection, so nothing lands on the canvas until you decide it should. Links pointing at a node that does not exist are flagged rather than quietly ignored, and the ones that deserve to be drawn are the next step.
Hold ALT and drag from one node onto another, or use the connect button in the bottom bar. Then give it a type, a direction, a label and a weight.
Dark Energy is our auto-arrange and spreads everything into clusters according to what pulls on what.
Layers sorts the same nodes into a hierarchy by type.
Switch to 2D for a flat, readable diagram, and switch back for the depth.
It is one structure seen from different vantages, never a second copy.
Your files never stop being files. Open one in another editor, save it, and the canvas updates while you watch. Hang a PDF, a spreadsheet or a drawing on a node and it opens in whatever you already use for it. They also stay readable by anything else that reads files, including whichever AI you already use.
Most tools can record that two things are related, which is a single bit of information and rarely the interesting one. Knowing that a person and a grant are connected tells you almost nothing, whereas knowing that she holds it, and that a dataset is key evidence for it, is the beginning of an actual picture.
So a connector here carries four things: a type you defined, a direction, a label in your own words, and a weight that decides how hard it pulls in the physics. In the space below, the arrows themselves read fine-tuned on, produced by, release-gate score and leading investigation, which is a structure making claims rather than drawing lines.
Node types work the same way, so that nodes become Clients, Risks, Sites, Sources, Witnesses or whatever your field genuinely calls them, each with its own colour on the canvas. The result is a structure that speaks your vocabulary back to you, which is the difference between a diagram you have to translate and one you can simply read.
A note is the right place for reasoning and the wrong place for a date, an owner or a reference number. Those want looking up rather than reading.
Define a property once for the whole space, in the sidebar beside your node and connector types, and it turns up on every node waiting to be filled in. They come in kinds rather than data types, so a date behaves like a date, a rating is out of three or five, and a choice comes from a list you curate or extend as you go.
There's a Key kind for the values you copy rather than read. It stays hidden until you reveal it, one click copies it, and it's kept out of search so a fragment can't bring the node up. It's marked with an eye and a line through it rather than a padlock, because hiding is all it does and nothing in your files is encrypted.
Every value is plain text in the node's own front matter, so anything you write is still readable by hand, in the file, without us.
You are not meant to keep one structure and feed it forever until it becomes an archive nobody can navigate. Make a space for the investigation, the client, the project or the year, and give each one a shape you can hold in your head.
When two of them need to meet, drop a Bridge between them and travel across with a double-click. Have as many spaces as you like and as many Bridges as you like on the free plan, permanently.
What you get is a distributed set of focussed structures joined in-situ, which is how most people's work is actually shaped. Each space holds up to twenty nodes on the free plan (roughly one finished piece of work), and Professional removes the ceiling entirely.
There is a longer argument underneath this, about why work resists being filed while it is still moving.
Fifty-six backgrounds for the canvas, seven themes for the interface around it, any image you care to upload, and a Pexels search so you do not have to go and find one first.
By the time you read this, it's probably more.
Observation
Two controls that matter when you choose an image are blur and dim. They exist so a background can be an image you actually like rather than a compromise, and the text stays readable. Customisation that costs you legibility is not a feature but a trade.
We add to the set fairly frequently, partly because people ask and partly because it is an enjoyable afternoon's work on a Friday.
The graph itself is adjustable in the same way. Pick fonts, size the cards and the connector labels, decide whether a node grows with the number of things it connects to, and cap how many nodes are drawn at once when a space gets large.
The way it arranges itself is adjustable too.
Pop a node out into its own window and put it where you want it. On two screens you can have multiple notes on one and the structure on the other. On one screen it is a note that can be Alt+tab'd between.
The argument is above, and this is the part that answers "yes, but does it actually do".
Every item on that list is in the free plan, and we are aware of how that reads as a business plan. Nothing is being made of you in the meantime, because there is no account and no telemetry to make it from. The part we charge for is on the publishing page, and it is the part that leaves your computer.
No account, no card, nothing to cancel. Every install opens on Professional for fifteen days, and what follows it is a free plan that does not expire. There is also a help section built into the app, with a practice space you can poke at without breaking anything of your own.