Hi HN, I'm Flynn. I built Arbor because literature review was the most painful part of my PhD.

Arbor takes a research question and builds a knowledge graph:

  1. Decomposes the question into 4 focused sub-inquiries
  2. Searches arXiv + Semantic Scholar in parallel
  3. Screens papers for relevance and extracts findings
  4. Synthesizes everything into structured markdown

All stages stream in real-time via SSE. The graph is interactive — click any node to see sources, findings, and connections.

Stack: React 19 + TypeScript + React Flow (frontend), FastAPI + SQLAlchemy + aiosqlite (backend), hosted on Vercel + Railway.

It's free, no signup, no login. Rate limited to 5 queries/hour per IP.

Limitations I'm aware of: only searches arXiv and Semantic Scholar (no PubMed, no Google Scholar), graph layout can get crowded on large queries, and LLM extraction quality varies.

I'd love feedback on: research quality, graph UX, missing features, anything. What would make this useful for your work?