Capacities is not end-to-end encrypted, and that is a deliberate choice. End-to-end encryption makes the server blind to your notes. Most of what makes Capacities useful today, from the AI assistant to connecting ChatGPT or Claude to your notes, depends on a server that can read your notes when you ask it to.

Your notes are encrypted while they travel between your device and our servers, and while they are stored on our disks. This page explains why we stop there, and what we do instead to keep your notes private and safe.

What end-to-end encryption means

With end-to-end encryption (often shortened to E2EE), your notes are locked on your device before they leave it, and only your devices hold the key. Our servers would store the locked notes but could never open them. Neither could we, nor anyone who broke into our servers.

That is a real and strong protection. We do not want to talk it down.

What people really want when they ask for it

When people tell us why they want end-to-end encryption, the answers are rarely technical. They sound like this:

• "My journal and my health notes are nobody's business."
• "A data breach is not a question of if, but when."
• "I don't want a company to sell my thoughts or train AI on them."
• "If the key isn't mine, I never really know what happens to my data."

Underneath these are a few simple wishes: privacy (nobody reads my notes), safety (my notes do not leak and do not get lost), respect (nobody profits from my notes) and control (I decide, and I can leave). End-to-end encryption answers the first two very well. It says little about the last two.

There is also a simpler reason. When a product page lists "end-to-end encrypted", it looks like a free extra. If you can have it, why would you say no? Most of us see it as one more tick on a checklist, like dark mode or a desktop app.

It is not a tick on a checklist. It is a decision about how the whole product is built. Once the server cannot read your notes, it cannot do anything with them for you. Every feature that needs to understand your notes either runs on your own device or does not exist. So the honest question is not "Do I want end-to-end encryption?" but "What am I willing to give up for it?"

AI needs to read your notes

This is the most important reason, and it matters more every month.

The AI assistant inside Capacities works with your whole space. It searches your notes, reads the ones that matter, follows the links between them and, when you approve, creates tasks or updates notes. With AI chat connectors, you can also connect ChatGPT, Claude, Cursor or Le Chat to a space, so the AI you already use can look through your notes and save results back. Connectors use an open standard called MCP (Model Context Protocol). The Capacities API works in a similar way for other apps and your own scripts.

All of this needs something that can read your notes. The AI models that do this well are very large and run in data centers, not on a phone or a laptop. Connector requests do not even come from your device. They come from the AI company's servers to ours, often while your laptop is closed and your phone is in your pocket. A server that cannot read your notes cannot answer them.

This is the part that is easy to miss: end-to-end encryption and cloud AI cannot both apply to the same note. In an end-to-end encrypted app, your device has to unlock a note and send the readable text to an AI provider. From that moment on, the note is no longer protected by end-to-end encryption. For connectors, the only way around this is to give the server your key. But a server that holds your key can read your notes, which is exactly what end-to-end encryption is meant to prevent. You would have the extra steps of end-to-end encryption without its protection.

AI models that run on your own device avoid this problem. Today they are far weaker than cloud models, and they cannot answer a request from ChatGPT while your devices are switched off.

We believe your notes will become the memory you bring to every AI tool you use, and we want that memory to be easy to connect. AI in Capacities is opt-in, and you can turn it off for individual spaces. When you use it, the AI providers we work with may not train on your notes or keep them. Read the details in AI and privacy.

Saving from anywhere

The same is true for capturing ideas. You can send a message to Capacities on WhatsApp or Telegram, forward an email, sync highlights from Readwise and Kindle, and connect your calendar. All of these services talk to our servers, not to your device, and none of them are end-to-end encrypted on their side.

To be fair, some of this could be rebuilt. A server could lock an incoming message the moment it arrives, so that only your device can open it later. But anything that needs to understand what arrives would have to wait until you open the app on one of your devices.

The quieter cost: a lost key means lost notes

With end-to-end encryption, the key belongs to you alone. That is the point, but it also means nobody can help you. If you forget your password and lose your recovery code, your notes are gone for good. The company cannot restore them.

For many people, safety means "my notes will still be there in ten years" at least as much as "no server can read them". In Capacities, losing your password does not mean losing your notes, and our daily backups in several European locations protect them from loss.

You trust a team either way

End-to-end encryption does not remove the need for trust. It moves it.

The encryption happens inside the app, and a company writes the app. With an end-to-end encrypted app, you trust that every update still locks your notes properly and never sends your key anywhere. Very few people can check that for themselves, for every update, on every device. You also still trust the company to keep the app running for years and to let you take your notes with you when you leave.

So the more useful questions are about the people behind an app:

• How do they earn money?
• Who owns the company, and could it be sold?
• Which laws apply to your notes?
• What happens to your notes if you leave, or if the company does?

An app can be end-to-end encrypted and still give weak answers to these questions. An app can also give strong answers without it.

Our answers

• How we earn money. Only from people who pay for Capacities. We never show ads and never sell data. We use your notes for one purpose: providing the service to you.
• Who owns us. Capacities Labs GmbH in Freiburg, Germany. We are bootstrapped and have no outside investors.
• Which laws apply. German and EU law. We give every user the rights of the GDPR, wherever they live, and your notes are stored on EU servers.
• AI. AI is opt-in, and you can limit it to the spaces you choose. Our AI providers may not train on your notes, and may not keep them except for limited abuse monitoring.
• Leaving. You can export everything to Markdown and other formats at any time, and schedule automatic exports to your own device. See Export.

How we keep your notes out of reach, even from us

In theory, whoever runs the servers could read what is stored on them, and we want to be open about that. So we set up Capacities to make it hard, including for ourselves:

• Your notes only go to your account. Capacities hands out your notes only to someone signed in with your login. A strict permission system, which we test regularly, makes sure of that. The only exception is content you decide to share, and then only people with the link can see it.
• Our hosting provider cannot read them. Your notes and files, including all backups, are stored on encrypted disks that only Capacities' own software can unlock.
• Your notes live on separate machines. The servers that store your notes are physically separate from the computers we use to build and test Capacities. Our everyday work happens nowhere near your data.
• Only one computer can reach those servers. Access requires a secret key that is itself protected by a password.
• Even with access, there is no window into your notes. The programs on those servers run in sealed containers that only talk to each other. Getting at anyone's notes directly would mean deliberately changing the whole setup, which takes real effort and expert knowledge.
• Links to your files expire. When you open an image or a PDF, you get a private link with a long secret code that stops working after 12 hours.

These barriers make our own work harder, and we built them on purpose. Reading your notes would mean breaking the rules we set for ourselves and destroying the trust our whole business depends on. You can read the full detail in our data protection memo.

Partial encryption for your most sensitive notes

If you need a guarantee that no company could ever read your notes, even in theory, then end-to-end encryption is a hard requirement for you, and we respect that.

We are thinking about partial encryption: letting you lock individual notes that you explicitly mark as sensitive, with a key only you hold. Those notes would be left out of search and AI features, while the rest of your space keeps working as it does today. This is an idea, not a promise. If it matters to you, tell us on our feedback board.

Until then, you can keep sensitive notes in a separate space with AI turned off. The AI features in Capacities never send content from that space to an AI provider, and no outside AI can reach it unless you connect that space yourself.