We built Pepiros because we kept seeing the same problem with research papers; people we know confidently summarise, cite, and discuss papers without checking what the paper actually says. AI makes this worse because a polished answer can sound correct even when it is not supported by the source.
We wanted to build a system where claims are tied directly to their source text, so you do not have to trust the person, model, or platform. You can check the evidence yourself.
Pepiros turns research papers into structured maps of their claims, evidence, figures, numbers, equations, references, and everything else one could want. It then lets you:
• Read a paper with its claims linked to the exact supporting text on the page. • Audit any external summary sentence by sentence against the paper. • Explore relationships between papers, claims, and findings as a graph. • Ask questions using only the indexed papers, with unsupported answers explicitly rejected. • Publish and discuss papers, with comments attached directly to specific claims rather than just posts. • Use the same verification system through an MCP server for Claude.
Built With
- anthropic-claude
- drizzle-orm
- esbuild
- framer-motion
- groq
- model-context-protocol
- next.js
- openalex
- pix2text
- postgresql
- pymupdf
- python
- radix-ui
- react
- react-flow
- react-pdf
- recharts
- semantic-scholar-api
- supabase
- tailwindcss
- tiptap
- typescript
- vercel
- vitest
Log in or sign up for Devpost to join the conversation.