Turns a research PDF into a living knowledge graph.

Most tools that summarise a paper ask you to trust the summary. The model produces confident prose, the prose sounds like the paper, and the only way to know whether it is right is to go and read the paper yourself, which is the work you were trying to avoid.

Pepiros takes the opposite position. Every claim it surfaces is matched against the exact sentence it cites, with a score you can see. Claims that clear the threshold show their quote, their page, and their citation id. Claims that do not are labelled inference and get no citation at all.

Why the badge never says “verified”

A fuzzy-matched quote proves quotation provenance. It does not prove entailment. A model can attach a real sentence from the Methods section to a conclusion that sentence does not support, and still score a perfect match.

So the badge reads quote located, never verified, and the claim and its quote render next to each other rather than one behind the other. The reader adjudicates whether the claim follows. The matcher only establishes that the sentence is really there, at that page, in that paper.

An entailment overlap floor helps at the margins: every number, unit and comparator in a claim also has to appear in the anchored span. That catches the failure a fuzzy match alone misses, a genuine quote attached to a reversed or overstated conclusion. It is a floor, not a guarantee, and it is described that way on how it works.

Built as a service, not just a site

The same grounding is callable over MCP, so an agent can check its own claims against a source before asserting them to you, and say so out loud when one comes back unsupported. That is the part we think is genuinely new, and it is documented on the agents page.

Who builds it

Pepiros is built by Anay Dhawan and Yash Kewlani, under the StudentSuite org. It started as a hackathon project and is still an early build, moving quickly. The roadmap says what is real today and what is not, without rounding up.