Deep research
Per-node, agent-driven research grounded in arXiv and the web.
Click any node and hit Research. An agent expands the node title into a search plan, calls Tavily for web sources and arXiv for academic papers, then synthesises an overview, key points, and follow-up questions.
Pipeline
- Plan — derive 3–5 sub-queries from the node title and surrounding context.
- Fetch — Tavily + arXiv in parallel; cache hits skip the call.
- Synthesise — overview, 5–8 key points, 3 follow-up questions.
- Persist — cached on the node so re-opening is instant.
Cost
A research run costs 1 token. Re-opening a previously researched node is free; click Re-run to force a fresh report.
Was this page helpful?