Your Questions Answered

An AI research assistant with a memory. You talk to it in plain language about your field, the people whose work you follow and the papers you are reading, and it keeps that context between conversations rather than starting from nothing each time.

It looks up researchers and papers and answers questions about them: a person's output over time, how two researchers' work overlaps, who cites whom, how a field has shifted across a decade. It keeps watchlists of people and papers you care about and ranks them the way you ask. It reads PDFs you point it at and answers questions about that document. It searches the web and Wikipedia when a question needs something outside the literature. And it remembers what you tell it, so you do not have to explain your work twice.

We do not train a model of our own. Acadex uses established language models from external providers and adds the parts that make them useful for research: the bibliographic data, the memory, and the reasoning that decides which records to look up before answering. Today the model answering you is provided by Mistral AI, a French company. The app shows you which model is in use, and our Privacy Policy explains exactly what is sent where.

From OpenAlex, the open catalogue of scholarly work, which is our primary source for papers, authors and citations. Coverage, metadata quality and disciplinary breadth follow OpenAlex's own. We do not claim to know about work that is not in the record. We are adding further open sources as we ingest them.

Almost certainly, since OpenAlex spans every discipline. How well it does depends on how completely your field is represented there: well-indexed, heavily-cited literature gives better answers than work that is sparsely catalogued. If your area looks thin, tell us, because that is useful to know.

The h-index combines how much someone has published with how often it is cited: an h-index of 20 means 20 papers with at least 20 citations each. Acadex can show it, along with related measures, and can plot how they have moved over time. It is one view among several, and a crude one. it is slow to move, it ignores authorship order, and it is not comparable across fields. We show it because people ask for it, not because we think it measures your worth.

There is a free tier with a monthly allowance, enough to try it properly. Paid plans raise that allowance. See Plans for what each one includes.

Your account and conversations are stored on servers in Germany. Message contents are encrypted in our database. To answer you, the relevant part of a conversation is sent to our AI provider in France and comes back as a reply, both inside the EU.

Your conversations are never used to train a model, ours or anyone else's. We do not sell them and we do not mine them for features.

One thing we state plainly rather than glossing: a searchable copy of each stored memory is kept in readable form, because the assistant has to be able to search its own memory to be useful. The full picture, including what that means, is in our Privacy Policy.

Yes, and we think this matters more than any feature on this page. The memory is not a black box: you can open it, read what it has recorded in plain language, correct anything it got wrong and delete anything you would rather it did not keep. Deleting your account removes your conversations, memories, watchlists and settings with it.

English and German, throughout: the interface, the answers, and what it remembers.

Write to info@acadex.io, or use the contact routes at the bottom of this page. We read all of it.