Why inveni?

Scientific knowledge grows exponentially. A researcher's capacity to follow it stays the same.

In 2026, arXiv releases one to two thousand papers every working day — how many of them did you read?

Ever missed a competitor's work? Had a second reviewer claim your article is not novel? In today's fast-paced science, missing one important piece of information can cost months of work.

Meet inveniI found (lat.)

inveni works as your personal academic analyst, helping you discover the research that matters. We train a model tailored to your research interests; it reads new publications every day and keeps you alerted.

But what if you need to find or verify specific information hidden somewhere among millions of published papers?

A plain search-engine query will not be efficient. AI has become a powerful research tool, outperforming a single scholar with a browser search bar. However, AI is quite limited in data retrieval, and it hallucinates when it has to rely on memory. And the cost of a wrong source or a missed critical article may be your reputation, a failed journal submission or a million-dollar grant interview.

inveni integrates with ChatGPT and Claude, grounding your AI in the literature, enabling efficient search and letting it analyse your inveni profile. You can call an inveni literature search right in the AI chat, or ask it to analyse your inveni recommendations.

arXiv submission volumes grow exponentially

New submissions per month, July 1991 – August 2026

31k
papers in August 2026
30k 20k 10k Aug 2026 · 31,173 papers
arXiv submissions per month, 1991–2026. Source: arxiv.org/stats.
How inveni works

Two judges for every paper, one short deck for you

Daily recommendations

We ask you to show us examples of the papers that describe your research, and we train your personal model on them. Every day it scans thousands of papers and discards about 98.6% of them, leaving only a few top picks. This sample is then scored by a frontier AI model. Both judgements — your personal model's and the AI's — are combined to decide whether to show you the paper, filtering out about 99.8% of all candidates. inveni covers the whole of arXiv and leading journals from Nature, Physical Review and Science.

Get your feed delivered in Telegram or by email shortly after the articles are published.

When you receive a paper, view the AI summary and figures to decide quickly whether it is worth a read. Give it a like or dislike to fine-tune your model. Share the article with your colleagues, or save it directly to your Zotero or inveni library.

Your personal AI research

AI is powerful — but only if it knows where to look and gets what it needs. Your chat cannot reach most of the papers on the web, stopped by robot blockers and API rate limits. When it runs out of evidence, it starts to rely on its memory and may invent non-existent references.

inveni helps AI cover its blind spots and gives it access to the resources it needs. Connect inveni to your ChatGPT or Claude and enable search over a 24-million-edge citation graph, semantic search, and direct full-text search in a database of almost 3 million arXiv articles — give your AI ground and optics.

Your AI assistant, alone

Limited retrieval

Rate limits, robot blocks, can't read PDFs

Knowledge cut-off

Relies on memory — and hallucinates references

Doesn't know your research

Generic answers, not tuned to your interests

Your AI assistant + inveni

Reliable retrieval

Sees every paper's full record — LaTeX full text, citations, DOI

Search

24M-link citation graph, semantic and keyword search, full-text search

Personal academic analytics

Catch up, digest, trends — on your own library

Who is behind inveni?

A word from the creator

I'm a physicist (PhD) at the National University of Singapore. I built inveni in my own time, on my own servers, because I needed it.

I never liked the chore of keeping up with papers, and I found that most of my peers struggle to follow the latest publications or get stuck under a pile of articles to read. It did not bother me until I started to face the consequences of not keeping up — I found out about a competitor's work after submitting my paper, a reviewer's response from Nature Physics claimed my work was not novel without citing a single reference, and I kept finding better methods in the literature months after pursuing inefficient ones.

When I realised how many people cannot keep up with the latest publications, I decided to create a dedicated tool to aid our research, keep us always aware, and make idea generation and paper reading easier and more enjoyable.

Born from this intention, inveni was developed with suggestions from tens of full-time researchers to bring maximum benefit to scientists.

My aim is to make inveni your main connection point with the literature — from daily personalised recommendations to grounded AI research on demand.

FAQ

Frequently asked questions

Can my Claude recommend papers on its own?
No, for two reasons. First, it will not reach all publications, missing most of what you might need. Second, pure AI does not get your taste. Read the experiment report.
What happens to my personal data?
Your email is used only for sign-in and the digests you turn on. API keys you connect (for example Zotero) are stored encrypted. We never sell your data or use it to train models for other users.
Do I get recommendations of paywalled journal publications?
Yes. We use publicly available metadata posted by the main publishers to score a paper's relevance for you.
What if I miss a few weeks?
You can find all your recommendations in your History. Too many recommendations? Ask your personal AI paired with inveni to cook up a digest for you (works on the free plans of ChatGPT and Claude).
Why should I use inveni over other recommendation services?
inveni is one of very few services that offer high-quality same-day recommendations. It stands out with a two-judge relevance scoring system that improves on your feedback. Furthermore, inveni is, to the best of our knowledge, the only service that offers AI integration with the full arXiv corpus, equipped with hybrid search tools.
Is inveni free?
Yes, inveni is free. Moreover, it lets you save on the AI tokens you spend on research by providing efficient infrastructure. inveni might become a paid subscription service in the future.
What fields are covered?
All of arXiv — physics, computer science, mathematics, statistics, quantitative biology, quantitative finance, economics and electrical engineering — plus journal feeds from the Nature family (Nature, Nature Physics, Nature Materials, Nature Communications, Nature Nanotechnology, Nature Photonics and more), the Physical Review journals (PRL, PRX, PRB, PRX Quantum) and Science and Science Advances.
How do I use inveni with Claude and ChatGPT?
Install the inveni connector in a few clicks, following the instructions in your profile. Then just ask your chat to use inveni alongside any query.