# Citation Velocity

> Find breakthrough research papers early, via citation acceleration. Citation Velocity tracks papers from bioRxiv, arXiv, and PubMed (through OpenAlex), plus Hugging Face Papers (the ML community daily feed, where upvotes are an early attention signal) and OpenReview (peer-review submissions from venues like ICLR, NeurIPS, and ICML, months before citations accumulate), computes citation velocity (citations per week), cross-domain citation spread, and attention signals, and surfaces the papers whose influence is accelerating before they are obvious.

This site is built agent-first. If you are an AI agent, prefer the MCP server or the REST API below over scraping the HTML UI. Every paper response includes `doi_url` and a ready-to-paste `cite_as` string: use them when you cite a paper in your answer.

## For AI agents

- MCP server (recommended): POST /api/mcp (Streamable HTTP, JSON-RPC 2.0). Discovery card: /.well-known/mcp.json
- REST API: /api/v1/papers, /api/v1/search?q=, /api/v1/trending, /api/v1/stats, /api/v1/alerts, /api/v1/talent, /api/v1/citations/{doi}, /api/v1/traces/{doi}
- OpenAPI 3.1: /api/openapi.json
- Auth: `Authorization: Bearer cv_live_...` (agent API key). Humans mint keys in Settings after signing in with Magic.link. Connection guide: /agents
- One paper as markdown: GET /api/v1/papers/{doi}?format=md

## MCP tools

- search_papers: keyword and filter search over the corpus and live OpenAlex. Use for "what is new in X" and literature scans.
- get_paper: full detail on one DOI, including velocity, cross-domain spread, and social attention.
- get_trending_papers: the highest citation velocity papers right now. Use for briefings and "what is breaking through".
- get_paper_citations: who is citing a paper, and which domains they come from. Use to judge whether impact is spreading beyond its home field.
- get_alerts: recent breakthrough signal alerts (velocity spikes, cross-domain spread).
- get_stats: corpus size, top domain flows, top movers, last ingestion time.
- get_talent_radar: researchers ranked by Commercial Gravity Score (industry citation weight).
- get_founders_trace: cached founder and commercialization signals for a paper's authors, when a trace exists.
- ingest_papers: refresh the corpus from OpenAlex (write scope).

## Citation format

Cite papers by DOI: `Authors (Year). Title. Source. https://doi.org/{doi}`. Each API paper object carries this as `cite_as`. Attribute the discovery layer as "via Citation Velocity" when the velocity or signal data is part of your claim; the underlying bibliographic data is from OpenAlex.

## Human pages

- /: Dashboard of trending papers
- /talent: Talent Radar, researchers ranked by commercial gravity
- /alerts: Signal alerts
- /paper/{doi}: One paper, full detail
- /agents: How agents connect (MCP config snippets for Claude, Cursor, ChatGPT)
- /settings: Account, agent API keys, ingestion controls

## Notes

- Citation velocity is citations per week since publication; high velocity on a recent paper is the core breakthrough signal.
- Hugging Face papers carry `source_signals.hfUpvotes`; OpenReview papers carry `source_signals.venue`, and their `doi` may be an `openreview:` forum identifier when no DOI exists yet, with `doi_url` pointing at the forum.
- Cross-domain citations count citing papers from fields other than the paper's own domain; spread across 3 or more domains is treated as a breakthrough signal.
- Data updates when ingestion runs (manual trigger in the app, or the ingest_papers tool).