Review the Whole Field, Not One Index at a Time

A sweep runs across OpenAlex, PubMed, arXiv, Crossref, and Semantic Scholar in one pass — plus code registries and model hubs like GitHub and Hugging Face when the question is about software health. That’s part of 500+ sources behind 3,500+ endpoints. If your field leans on one index, you can scope the ask to just that one.

The schedule lives in your agent, not in Anysite. Set a scheduled task in Claude — or any agent that can run on a timer — and on each run the agent pulls fresh records from Anysite over MCP, then handles delivery itself: filing the week’s papers into a Sheet, posting preprint alerts to Slack. Anysite is the live data layer; your agent is the scheduler.

Every record is fetched live at query time from the index itself, and each result traces back to the fetched record — no cached copies going stale. Honest caveat: the indexes disagree with each other on citation counts, so treat them as a ranking signal, and check which index a figure came from before it goes into the paper.

No — and it shouldn’t. Anysite gets your agent the records: papers, citations, release histories, pulled live. Screening, inclusion criteria, and synthesis stay yours, and the results file into whatever Sheet or reference workflow your review already uses. It removes the index-hopping, not the scholarship.

MCP plans are flat from $30/month, with a 7-day free trial — enough to run a real review end to end before paying anything. A review is conversational: one sweep across the indexes, then follow-up asks as you walk the citations — and a heavy review month stays one flat plan, not a stack of per-index subscriptions. Want the same catalog inside your own code? REST and CLI come on one API key, usage-based from $49/month.