Peer Research Papers MCP
Search nine scholarly sources in one call, screen any paper for retractions before you cite it, find a legal free copy, walk citation graphs, and track a question so you only ever see what is new. No API key.
How to pay
Subscribe
$19/month
Predictable monthly cost with included usage. Best for steady, high-volume traffic.
- Unlimited tools within plan limits
- One API key, billed once a month
- Cancel any time
Literature search across nine independent sources, with retraction screening, legal full-text lookup, citation-graph traversal, and saved searches that report only what changed. No API key is required for any source.
Sources
| Source | Coverage |
|---|---|
| OpenAlex | 250M+ works, citation graph, open-access status |
| Crossref | Registered DOI metadata for journals, conferences, books |
| arXiv | Preprints in physics, maths, CS, quantitative biology |
| Semantic Scholar | Paper graph, citation counts, open-access PDFs |
| Europe PMC | 43M+ life-science records, incl. PubMed and preprint servers |
| PubMed | Biomedical citations curated by the US National Library of Medicine |
| DOAJ | Articles from vetted open-access journals |
| Crossref Crossmark | Retractions, expressions of concern, corrections |
| Unpaywall | Legal free copies, with licence and version |
Check before you cite
integrity_check reads publisher-asserted Crossmark updates and tells you whether a paper has been retracted, corrected, or placed under an expression of concern. It correctly flags the Wakefield MMR paper and the Lancet Surgisphere hydroxychloroquine paper — the latter also under an expression of concern.
check_retractions screens an entire reading list in one call. Hand it the DOIs from a search and find out which ones you should not be citing.
A clean result is reported as "no retraction or concern found" — never as proof the work is sound. It means no publisher has deposited a notice, and nothing more.
find_full_text resolves a DOI to a lawful free copy via Unpaywall, with the licence and version attached, and flags "bronze" OA as free-to-read with no reuse licence.
Track a question over time
watch_query returns only papers you have not already been shown. Ask it monthly and you get a literature-review delta, not the same top-20 again.
The server is stateless: the cursor encodes what you have already seen, so nothing about you is stored server-side. The first call sets a baseline and deliberately reports nothing new. Every later call issues the same underlying search, newest-first, so a paper is never reported as new merely because the query drifted — the failure mode that makes most alerts useless.
Follow the citations
citation_graph walks both directions around a paper — what it cites, and what cites it, most-cited first — from a DOI, an OpenAlex id, or a URL.
literature_search fans out across the search backends in one call, grouped by evidence type.
Provenance is part of the answer
Every response names the sources behind it, with the caveats that matter for weighing evidence:
- arXiv results are preprints and NOT peer reviewed unless a journal reference is given
- Crossref holds registration metadata, not full text
- Semantic Scholar's anonymous API rate-limits hard; a gap means throttling, not absence
- PubMed indexing lags publication
- DOAJ vets journals, not individual articles
Call list_sources for the full register.
Limitations
This is retrieval and screening, not evidence appraisal. It surfaces papers, citation counts and publisher-deposited notices; it does not judge study quality or rank clinical relevance. Citation counts reflect attention, not correctness.