How it works
You write. Oknola files. Your agent reads.
One folder of plain markdown on a disk you own. Oknola types, tags and files what lands in it, keeps the structure from rotting while you work, and hands an AI agent only the tags you allow.
The mechanism
Four moves, and one of them is yours.
You capture. Oknola does the other three, on a schedule, whether or not you had a good week.
Start where you are.
Point Oknola at a folder you already keep, or start an empty one. Nothing is imported into a database, because there is no database. The files stay files, on your own disk, in the editor you like.
Oknola files what lands.
A note, a receipt, a client, a meeting: each one becomes a markdown file carrying its own fields, following OKF, the open markdown spec Google introduced in 2026. The app writes the type, the tags and the description, so you are not the one sitting there tagging.
The structure holds while you work.
Indexed Brain, our architecture, decides where a file belongs at 40 notes and at 34,000. A file that breaks the schema is fixed or refused, and that runs in both directions: what an agent writes has to conform too. Plain text all the way down, so every change is inspectable and reversible.
An agent sees what you point it at.
Tags are the permission model, not folders. A business agent gets your clients and your ledger; your journal stays dark, whatever folder it sits in. Agent off, your files never leave your device. Agent on, you chose the scope.
OKF
standardizes the file
Indexed Brain
organizes the files
Oknola
keeps both true while you work
What it looks like
The folder, drawn as the way you work.
Because every file is typed and tagged, the app can draw the business back to you: the path from inquiry to paid, what a client is worth, what is stalling and for how long. None of that is a second system to fill in. It is the same files, read differently.
The demo is the product truth, running on sample data. Click anything.
Go deeper
Each piece has its own page.
The folder
The Core
The free app: the structure, the format and the pass that keeps them true.
Read it →The layer
AI agent
Why context, not intelligence, is the bottleneck, and how the boundary works.
Read it →Your tools
Connect to Claude
Claude, Codex, Cursor or a harness on your own server, with nothing to install.
Read it →Native on macOS and Android first, iOS later. The structure, the pass and scoped agents work today; Operating Maps, Finance and Actions are in build, and the demo shows them working on sample data.
Waitlist
Start with the folder.
The free app comes first, on macOS and Android. Early access opens in waves.