Why we built a zero-knowledge company OS

A founder's company doesn't live in one place. The Stripe key is in a password manager. The roadmap is in one doc, the tasks in another tool, the customer data in someone else's cloud. Every one of those products wants your data on their servers, and most of them would love to train on it.

We think that's backwards. The most capable AI in history is here, and the price of using it shouldn't be handing over your secrets.

So we built lodos: a local-first, zero-knowledge company OS. One app on your Mac that holds your vault, your wiki, your tasks and a built-in AI agent, and a hard wall between that agent and your credentials. It can use your Stripe key. It can never read it.

What "zero-knowledge" actually means

Most tools sell you a security policy: a promise, written by their legal team, that they won't look at your data. Zero-knowledge replaces the promise with math.

In lodos, your data is encrypted on your device, with a key derived from your master password, before it is ever written to disk. The encryption uses standard, audited primitives: Argon2id for key derivation, XChaCha20-Poly1305 via libsodium for the actual encryption. We never see your password, never hold your key, and never receive anything but opaque ciphertext.

That has a blunt consequence we're happy to state plainly:

Not even we can read your vault. If you forget your master password and lose your recovery key, your data is gone, and that's the point.

It's the most honest thing a company can show you: a short list of what we can see (your email, your billing status, that an encrypted backup exists) and a much longer list of what we can never see (your secrets, your wiki, your tasks, your AI conversations).

Four ideas, no compromises

lodos is built on four commitments that we refuse to trade against each other:

• Local-first. Everything lives on your machine and works fully offline. The cloud is optional and never required.
• Zero-knowledge. Your data is ciphertext before it touches disk. Any server, ours or anyone's, only ever sees noise.
• AI-native. A built-in agent reads, plans and executes across your company with full context, and none of your keys.
• Yours. No lock-in, no tracking, no account required to download. Your company OS belongs to you.

You'll notice none of these is a feature you toggle. They're architecture. That's deliberate: a guarantee you can switch off is a guarantee that will eventually get switched off by accident.

The value is also in what it refuses to do

The most important features of lodos are the ones it will never have. There is no tool that returns a plaintext secret. There is no command the AI can run on your machine. There is no third-party model we proxy on your behalf.

And we wrote the build to fail if anyone breaks that promise. A compile-time invariant scan runs on every release: if a tool that leaks a secret ever sneaks in, the release simply doesn't ship. Refusal isn't a marketing line here; it's enforced in CI, so the guarantee can't quietly decay into a policy.

If you want the full picture of where the plaintext stops, read how we built the security model. If you want to understand the AI wall specifically, we wrote a whole piece on how your AI uses your Stripe key without ever seeing it.

Who this is for

lodos is built for the solo founder who wears every hat. Your Stripe keys, your AWS root, your customer data and your decision records all live in one place, and the AI takes the execution off your plate while you keep the strategy and the final call.

You keep the judgment. The AI takes the busywork. Your company stays yours.

Download lodos. It's free, local, and there's no account to create.