AI agents

Give your agent the whole business.

Agents are already capable. What they're missing is context. Oknola's structure is context an agent can actually use — every file typed, tagged and organized, in plain files it already knows how to read.

The problem

Your agent is smart. It just can't see.

Point the best model in the world at your business and it gets one chat window and whatever you paste into it. The real state of your operation — clients, offers, money, commitments — is scattered across apps agents can't read, or lives in your head, which they can't read either.

The bottleneck stopped being intelligence. It's context.

Why plain files

A folder is the universal interface.

Every serious agent reads and writes files natively — it's the one interface they all share. No API keys, no integration project, no waiting for an official connector.

If your business is a folder of plain text, any agent can work on it today — and any agent released next year can too. That's why Oknola keeps nothing in a database: the files are the source of truth, and the files are the interface.

agent · ~/business

$ agent ask "which client is stalling?"
› reading 03-clients/ … 41 files
› cross-checking 05-finance/invoices
Northwind — proposal sent 19 days ago,
no reply, invoice #204 still open.
✓ follow-up drafted → 03-clients/northwind

What the structure adds

Readable isn't the same as understood.

A folder of random notes is readable, not legible. Three things turn Oknola's folder into something an agent can genuinely operate on:

Every file follows OKF.

The open, vendor-neutral markdown standard Google introduced in 2026, designed precisely for handing agents curated context. Typed, tagged, described: the agent knows what a file is before reading a word of it.

Organized by Indexed Brain.

Our architecture. Predictable structure at any scale, so an agent finds your finances, your clients, your commitments where the structure says they are. Navigation, not guesswork.

Oknola enforces both.

So the structure never rots. Stale structure is how you get a confidently wrong agent; enforcement is how you get one that stays right.

OKF

standardizes the file

Indexed Brain

organizes the files

Oknola

keeps both true while you work

Bring your own agent

Works with the agent you already trust.

Claude, Codex, OpenClaw, Hermes — if it can read a folder, it can work on your business. On your desktop: point it at the structure and ask. On a server: run a 24/7 harness on a VPS you own, working the same synced folder while you sleep.

We don't pick your agent. Compatibility isn't a partnership we signed — it's a property of plain files.

Desktop

Claude · Codex

Point it at the structure and ask.

Server, 24/7

OpenClaw · Hermes

A harness on a VPS you own, on the same synced folder.

No MCP server. No plugin. No connector to wait for.

The boundary

It sees what you point it at.

Business and personal live in separate areas of one structure — and that separation is the permission model. An agent gets the area you give it, nothing more: your pipeline without your journal, your training log without your client files.

And the two-state doctrine holds always — agent off, your files never leave your device; agent on, you chose the scope.

Where this sits on the roadmap

True today. Bigger later.

Everything above about bringing your own agent is true of OKF plain files right now — it's how the founder runs his own business today. If you want an agent on your business this week, the setup-on-your-own-VPS service is live.

Further out sits the Oknola agent layer: full orchestration — an agent that holds your entire business context and works across every module. In order, not on dates; the layers underneath ship first.

See the Roadmap

Waitlist

Put an agent on your business.

Want one running this week? The agent setup on your own server is live today — see Services.

Join the waitlist

No spam. Build notes and early access, nothing else.