Frequently asked questions

Their memory lives with the provider, works only inside their service and shows you only what the service decided to keep. Switch to another model, and the memory stays behind.

In MindDuck the memory is yours: all your chats, documents and facts live on your devices, and any model answers with them at hand. You can open, correct or delete every fact, and every answer shows where it came from. Switch models as often as every message without losing anything.

About 10 minutes: install the app, paste a model key and, if you like, load your ChatGPT or Claude archive. Every step is in Getting started, and how to get a key is in Connect a model.

No subscription needed: you connect your own key and pay the provider only for what you use.

Affordable models such as Gemini 3.8 Flash or DeepSeek 4.1 Flash, with memory across all your chats, handle everyday tasks very well. Through OpenRouter that usually comes to under 10 euros a month. For the occasional hard question you can always switch to the strongest model: such an answer costs more, up to 50 cents - 1 dollar, but you pay only for that one.

With a local model in Ollama or LM Studio you can pay nothing at all, and then answers don't need the internet either.

The final price isn't set yet. The plan is 29 to 49 euros for a license covering three synced devices.

A purchased version keeps working on your devices forever, and it gets updates for a year. After that you can renew updates or simply keep using the version you have. Beta participants will get a special price.

We see none of the content. We have no server where your chats could end up; the site knows only your account email and your list of devices.

The model provider receives your message and the pieces of memory found for the answer, not your whole base. Under their API terms, major providers generally don't train models on requests. Check your own provider's terms: on OpenRouter some free models follow different rules.

Google Drive, if sync is on, stores only encrypted packages that can't be read.

Full privacy comes with a local model: then nothing leaves your device. What goes where, in detail: Data and privacy.

If you saved your recovery code, install MindDuck on a new device and your workspace is restored from the encrypted copy on your Google Drive.

Without the code and without your devices, nobody can read the data - neither you nor us. That's the other side of the key being only yours. So save the code right after you set up sync.

The stable version keeps working: access is checked on the device, without our server. Your data lives on your devices and in your Google Drive, not with us.

You can export everything to HTML at any time - plain pages that open in any browser without MindDuck.

Download the archive from your ChatGPT or Claude settings and import it into MindDuck. The chats show up right away, with their original dates.

Then MindDuck builds the search index in the background. Even on a computer this takes from 10 to 40 minutes or more, depending on the size. You can use the app meanwhile; the imported chats become fully available for search and answers once indexing finishes.

You can keep using ChatGPT as usual and load a fresh archive later: chats already brought over won't be duplicated.

Yes. Index a large archive on a computer: it's a one-time heavy job. The finished base syncs to the phone, and the phone works with it just fine. Tested on 6.5 million words: over 2,000 chats and hundreds of documents.

Yes, you can attach PDFs already: MindDuck takes the text from them. Other formats will follow later. More in Attachments.

Not yet. MindDuck is a personal assistant, made for one person with several devices. If two people edit the same document or write in the same chat at once, sync can't reconcile them cleanly.

For shared work we plan a separate storage and search engine, including for companies and multiple accounts.

Windows, macOS, Linux and Android, with iOS a bit later. One app and one memory on all your devices. Download.

Not yet. For now we're focused on making MindDuck a sustainable project that can keep growing for years. Your data doesn't depend on that: it stays with you and exports in an open format.

Programmers have "rubber duck debugging": explain your problem to a toy duck, and the answer often shows up by itself. Our duck listens too, except it also answers - and remembers everything you've told it.