Every filing system asks the same question at the worst possible moment. Where does this go? Filamental is built for the part of the work that can’t answer yet.
You answer the question at the beginning, when the work has barely started and you know less about it than you ever will again, and that guess becomes the shape. Six weeks later the project has moved somewhere nobody expected, and the folder it lives in is a record of what you thought on day one.
And filing is an act of hiding, which is what it is for. A tidy system is one where nearly everything is out of sight, and out of sight is indistinguishable from gone for anything you haven’t already decided matters.
You can’t weigh what you can’t see. The decision you need to make is usually about the thing you filed and forgot, and it’s missing from the reckoning, not because it was unimportant but because it was tidy.
Sort by how live a thing is, not by what it is about. That is a genuinely good idea, and it is why the PARA framework spread as far as it has.
Then it hands you four folders to do it in.
Actionability isn’t a place but a state, and the state changes. A live project isn’t a folder-shaped object: it’s a set of things pulling on each other, and several of them belong to three other projects at the same time. So you file, and then you refile, and the refiling is manual, and one week you quietly stop. That isn’t a discipline problem. The container is being asked to hold something it can’t hold.
It gets worse with other people. A folder tree is one person's opinion, and the only way in is to learn how that person thinks. What is obvious to you is obstructive to everyone else, and they probably won't say anything, just make their own copy.
One thing worth saying plainly. There’s an excellent and beloved Markdown notes application open on a second screen here most days, and nothing on this page is aimed at it. For knowledge that has been worked out and written down, it’s the best thing we know.
That’s a different job from the one described here, and Filamental doesn’t try to take it.
Written systems need you to write before they can hold anything. Something has to become a note first. Most of a live project never becomes a note: a supplier, a deadline, a decision nobody has made yet, a quote that’s already out of date, the person who has to sign. All real, all load-bearing, none of them documents.
In a space they’re nodes. A node can be a name, a position and three connections with nothing written in it at all, and it is still a full citizen of the structure. It means something through where it sits and what it touches, before it carries a single word. You can put an idea in front of you before you can explain it.
In a written system that’s an empty note and a broken link, something your weekly review tells you to clean up.
None of this is scaffolding you throw away. A space is a document, read differently from a page and for some things read better: who is involved, what depends on what, where the pressure is. That's why publishing one works. You send somebody a space and they don’t need your folder logic, your abbreviations, or half an hour of your time. The structure is on the surface.
There’s no export, because there was never an import. Every note in a space is an ordinary Markdown file in an ordinary folder. The spatial layer is one hidden folder alongside them holding positions, types and colours. Your notes stay exactly where they are.
Which is the point where the back half of PARA finally works, because by then you know what the thing was. You’re filing something finished instead of guessing about something in motion, and filing it wherever you already trust to keep it.
By then it’s ready for something else too. The structure is written down rather than implied: typed entities, named relationships, direction. Any MCP client, Claude included, can walk it and answer questions about it, because the answer sits in the files and not in the way you named your folders. Most archives are unreadable to a machine, since their organising logic lives only in a convention their owner never wrote down. This one doesn’t have that problem.
Filing is what you do once you know. A space is where the work lives until you do.
No, and it isn’t trying to be. A notes application is built for knowledge that has been worked out and written down, and that’s a job it does well. Filamental is for the stage before that, when the pieces are known but the shape isn’t, and most of them aren’t documents yet. The two sit next to each other because they hold different halves of the same work.
PARA is right about the thing that matters most, which is that you should organise by how live something is rather than by what subject it belongs to. Where it strains is the container. Four folders ask you to commit a thing to one place while it is still moving between all four, so the sorting has to be redone by hand every time the work changes. The idea holds, and the folder is the part that gives way.
Nothing has to happen, because there’s nothing to convert. Every entity in a space is already an ordinary Markdown file in an ordinary folder on your own disk, and the spatial layer is one hidden folder sitting alongside them holding positions, types and colours. File the folder wherever you file finished work. Your notes stay exactly where they are.
Yes, and more usefully than it can read most archives. A space writes its structure down rather than implying it: entities carry a type, relationships are named, and connections have a direction. Any MCP client, Claude included, can walk that structure and answer questions about it, because the answer is in the files rather than in a folder-naming convention only you understand.
No account, no card, nothing to cancel. Point it at a folder and start putting things down before you know where they go, which is the only time it matters.