Find the Papers Your Library Is Missing
Describe what you are working on, or point at papers you already have, and an AI agent searches the large scholarly catalogues for you. Every suggestion comes with a short reason why it fits.
Discovery Process in Action
- Write a request in plain words and optionally add example papers or a whole folder from your library.
- The agent searches OpenAlex, Semantic Scholar, Europe PMC (PubMed), arXiv and HAL, follows references and citations, and asks for recommendations.
- Suggestions appear as it works, and the run ends with a summary.
- Reply to steer it, for example “only after 2020” or “more on the methods side”. New papers are added at the top and ones that no longer fit are hidden.
- Every discovery is saved with its conversation, so you can come back and continue.
No invented references
The agent hands over each suggestion as a catalogue identifier, and the app looks it up itself before showing it. Anything that does not resolve to a real record is turned away, papers you already have are dropped, and a preprint is merged with its published version.
The workflow is designed so that you can trust what the agent suggests, with hallucinated papers filtered out.
Daily feed of new papers
Turn on Check for New Papers Daily in Settings.
- Collects recommendations based on your newest papers and new work citing papers you have.
- Ranks candidates on your computer by how close they are to your library and keeps the best matches.
- Follow authors, journals, arXiv categories and any RSS or Atom feed, so you don’t miss a publication from your favourite sources.
- Flags papers in your library that were retracted, withdrawn or corrected. We want to help you keep your library clean.
Gaps in your own library
Each paper lists the works it cites and the works citing it. The library graph shows works cited by two or more of your papers that you do not have yet, and one click adds them.
Which AI runs it
Claude Code, Codex, or any model reached by API key: Claude, OpenAI, Gemini, OpenRouter, or a local model in Ollama with tool support. The agent can only use the app’s search tools, with no access to your shell, files or the web.