Be the first to get access to new connectors. Get early access →
ConnectorBay Early access

Scholar — connector documentation

  • MCP endpoint: https://scholar.connectorbay.com/mcp
  • REST API: POST https://scholar.connectorbay.com/v1/tools/{tool_name}, described by the OpenAPI 3.1 spec at https://scholar.connectorbay.com/openapi.json
  • Transport: Streamable HTTP, stateless, MCP specification 2025-06-18 or later
  • Authentication: none required. An optional Authorization: Bearer <key> raises rate limits.
  • Health: https://scholar.connectorbay.com/health · Tool schemas: https://scholar.connectorbay.com/tools.json
  • Publisher: Unify Dynamics LLC · connectorbay.com/scholar · agent@connectorbay.com

What Scholar is for

Real references instead of plausible ones. When an assistant is asked for papers on a topic, the citation for a DOI, what is new on arXiv in a field, the latest trials of a drug, or whether a paper has a free PDF, Scholar answers from the indexes that hold the record — Crossref, arXiv, PubMed, Europe PMC, Unpaywall and ORCID — and returns DOIs and links so every claim can be checked.

What it does

Search across 170M+ Crossref records by topic and author with year and type filters, sorted by relevance, citations or date, with abstracts where publishers deposit them. One paper in full: metadata, ORCIDs, citation and reference counts, reference DOIs, funders, license, open-access status and best free copy, and a tl;dr when Semantic Scholar has one. Citations formatted by the DOI registry in APA, MLA, Chicago, Harvard, IEEE, Vancouver, Nature, ACM, AMA or BibTeX. arXiv search by words and category, the newest submissions in a category, and any paper by id with abstract and PDF. PubMed with full query syntax and structured abstracts, MeSH terms and free full-text links; Europe PMC with preprints, citation counts and open-access flags. Authors via ORCID (identity, institutions, works) and Crossref (most-cited). Journals by name or ISSN.

What it is not

Scholar does not summarise or judge papers; it returns what the indexes say. Citation counts differ by index. Abstracts and full texts belong to their publishers and are returned only where the index provides them. Semantic Scholar and OpenAlex are optional extras (keys raise limits) and never required.

Tools

Papers

search_papers — Search the scholarly record (Crossref: journals, conferences, books, preprints, datasets) by topic and/or author, with year and type filters, sorted by relevance, citation count or date. Each result has title, authors, year, venue, DOI, type, citation count and abstract when available. Use for "papers on X since 2020", "most-cited work by Y", "recent preprints about Z". Args: query, author, from_year, until_year, sort, work_type, max_results.

paper — Everything about one paper from its DOI: full Crossref metadata (authors with ORCIDs, venue, volume/issue/pages, type, citation count, abstract, license, funders, reference DOIs), open-access status and free PDF (Unpaywall), and a tl;dr plus influential-citation count from Semantic Scholar when available. Use for "tell me about this DOI", "who wrote 10.1038/nature14539", "is there a free version". Args: doi.

cite — A formatted reference for a DOI in a named style (APA, MLA, Chicago author-date, Harvard, IEEE, Vancouver, Nature, ACM, AMA) or a BibTeX entry — produced by the DOI registry's own citation formatter, so it is exact. Use for "cite this in APA", "give me the BibTeX for 10.1038/nature14539". Args: doi, style.

arXiv

arxiv_search — Search arXiv preprints by words and/or category, sorted by relevance or date — or, with only a category and sort='newest', the latest submissions in that field. Each result has arXiv id, title, authors, dates, abstract, categories, DOI/journal reference when published, and PDF link. Use for "recent arXiv papers on diffusion models", "what's new in cs.CL today", "arXiv search: quantum error correction". Args: query, category, sort, max_results.

arxiv_paper — One arXiv paper by id: title, authors, abstract, submission/update dates, categories, comments (pages, venue), journal reference and DOI if published, PDF link. Use for "what is arXiv 1706.03762", "summarise this arXiv paper". Args: arxiv_id.

Biomedical

pubmed_search — Search PubMed (37M+ biomedical records) with full PubMed query syntax: title, authors, journal, date, DOI, PMCID (free full text) and publication types (e.g. randomized controlled trial, review, meta-analysis). Use for "latest trials of semaglutide", "reviews on gut microbiome and depression since 2022". Args: query, sort, min_year, max_results.

pubmed_article — A PubMed article in full: structured abstract, authors with affiliations, journal, year, DOI, PMCID with free full-text link, MeSH terms, keywords and publication types. Use after pubmed_search or for "abstract of PMID X". Args: pmid.

europepmc_search — Search Europe PMC — PubMed plus preprints, Agricola and patents — with citation counts and open-access flags, sortable by citations. Use for "most-cited papers on long covid", "preprints about X this month". Args: query, sort, include_preprints, max_results.

Access, people, venues

open_access — Is there a legal free copy? Unpaywall's open-access status (gold/green/hybrid/bronze/closed), the best free PDF or landing page, its host (publisher or repository), version (published/accepted/submitted) and license, plus other locations. Use for "free PDF of this paper", "is this open access". Args: doi.

author — An author's identity and output: ORCID records matching the name (with institutions), the works listed on the chosen ORCID record, and their most-cited papers from Crossref. Use for "what has Geoffrey Hinton published", "find the ORCID for Jane Smith at MIT". Args: name, orcid, max_works.

journal — A journal's registry record (Crossref): title, publisher, ISSNs, subjects, total articles registered and output per recent year. Use for "which journal is ISSN 1476-4687", "how much does Nature Medicine publish a year". Args: query.

Conventions

  • Read-only. Every tool reads public data or computes; nothing is written anywhere and nothing about the user is stored.
  • Result envelope. ok: true with data, or ok: false with a structured error (bad_input, not_found, upstream_error). degraded lists optional sources that were unreachable.
  • Rate limits. Free tier 60 tool calls/minute and 300/day per user (keyed on the user identifier supplied by the host agent, else client IP); Pro keys raise this. Over-limit calls return HTTP 429 with Retry-After. initialize and tools/list are never counted.

Support

agent@connectorbay.com · connectorbay.com/scholar/support