Scholar Sidekick (REST API) — Citations, Retraction & Open-Access

Turn a scholarly identifier into a formatted citation, a bibliography file, or an integrity check (retraction / open-access / fabrication), via a documented REST API. No API key and no install required — plain HTTPS calls over curl. An optional free first-party key (ssk_…, from /account) or RapidAPI key only raises rate limits.

Prefer the scholar-sidekick-mcp skill instead if your host already has the Scholar Sidekick MCP server connected — same capabilities as native tool calls. This skill is the zero-setup path that works in any agent that can run curl.

When to Use

• The user has an identifier (DOI, PMID, PMCID, ISBN, arXiv, ISSN, ADS bibcode, WHO IRIS URL) and wants metadata, a formatted citation, or a bibliography file.
• "Cite this in APA/Vancouver/Chicago…", "give me a BibTeX/RIS file", "export these refs".
• "Has this been retracted?", "is this open access?", "is this citation real / did you make it up?"
• "Audit my bibliography", "check all these references at once", "which of these are fake or retracted?" — a whole reference list or .bib / .ris / CSL-JSON file.
• Do NOT use to search for papers by topic — that's discovery (see the arxiv skill). This assumes you already have an identifier.

Surfaces — call the API, never scrape the UI

The site is built for agents. The contract lives at:

• https://scholar-sidekick.com/llms.txt (index of agent surfaces)
• https://scholar-sidekick.com/AGENTS.md (REST + MCP guide)
• https://scholar-sidekick.com/openapi/openapi.yml (OpenAPI 3.1)

Always call the JSON REST API below. Do not drive the website form.

Authentication & limits

Calls to scholar-sidekick.com/api/* work anonymously — no key required — at a rate-limited free tier that is plenty for normal, human-driven agent use. For higher limits, create a free first-party API key (prefixed ssk_) at https://scholar-sidekick.com/account and send it as Authorization: Bearer ssk_…. Scholar Sidekick is also offered on RapidAPI for paid/managed tiers: subscribe at https://rapidapi.com/scholar-sidekick-scholar-sidekick-api/api/scholar-sidekick and call it through the RapidAPI gateway with your X-RapidAPI-Key. Default to the anonymous scholar-sidekick.com endpoints; add a Bearer ssk_… key for higher limits, or move to RapidAPI for paid volume.

Quick Reference

Base URL: https://scholar-sidekick.com

Procedure

Format a citation

curl -sS -X POST "https://scholar-sidekick.com/api/format" \
  -H "Content-Type: application/json" \
  -d '{"text": "10.1038/nphys1170", "style": "vancouver", "output": "text"}'

• text: one identifier, or several newline-separated for a batch. Pass verbatim — PMID:, arXiv:, ISBN hyphens, and https://doi.org/… are all tolerated.
• style: vancouver (default), ama, apa, ieee, cse, or any CSL style ID (chicago-author-date, harvard-cite-them-right, modern-language-association, nature, bmj, the-lancet, …).
• output: text or json. Response: { "ok": true, "items": [{ "formatted": "…" }], "text": "…" }.

Export a bibliography file

curl -sS -X POST "https://scholar-sidekick.com/api/export" \
  -H "Content-Type: application/json" \
  -d '{"text": "10.1038/nphys1170\nPMID:30049270", "format": "bibtex"}' \
  -o refs.bib

• format: bibtex, ris, csl-json, endnote-xml, refworks, nbib, rdf, csv, txt.

Check retraction

curl -sS -X POST "https://scholar-sidekick.com/api/retraction-check" \
  -H "Content-Type: application/json" \
  -d '{"id": "10.1016/S0140-6736(97)11096-0"}'

Returns { ok, doi, result: { isRetracted, hasCorrections, hasConcern, notices[], title } } (Crossref + Retraction Watch). One identifier per call — field is id. When the work has no DOI (e.g. a book), result is null and reason explains why (no_doi / timeout / upstream).

Check open access

curl -sS -X POST "https://scholar-sidekick.com/api/oa-check" \
  -H "Content-Type: application/json" \
  -d '{"id": "10.1371/journal.pone.0173664"}'

Returns { ok, doi, result: { isOa, oaStatus, bestLocation: {url, hostType, license, version}, locations[] } } (Unpaywall). One identifier per call — field is id.

Verify a claimed citation (catch fabrication)

curl -sS -X POST "https://scholar-sidekick.com/api/verify" \
  -H "Content-Type: application/json" \
  -d '{"claimed": {"title": "The title exactly as cited", "doi": "10.xxxx/xxxxx"}}'

Citation fields go inside a claimed object: title (required) plus an identifier (doi, pmid, … — recommended) and optional authors / year / container. Returns { ok, verdict, confidence, matched }, verdict ∈ matched / mismatch / ambiguous / not_found:

• matched — the claim agrees with the record at the identifier.
• mismatch — the identifier resolves but the title doesn't: the dominant AI-fabrication pattern (real DOI + invented title; Topaz et al., Lancet 2026).
• ambiguous — the identifier resolves to one paper but the claimed title matches a different real paper (a wrong-identifier error, not a fabrication).
• not_found — neither identifier nor title resolves anywhere.

Use this for "is this citation real?", not a plain format/resolve.

Audit a whole bibliography (batch fabrication + retraction check)

curl -sS -X POST "https://scholar-sidekick.com/api/audit" \
  -H "Content-Type: application/json" \
  -d '{"bibliography":"@article{ligo, title={Observation of Gravitational Waves from a Binary Black Hole Merger}, doi={10.1103/PhysRevLett.116.061102}}\n@article{fab, title={A Completely Invented Title That Does Not Match The Paper}, doi={10.1016/j.neuroscience.2023.02.008}}"}'

The batch counterpart to /api/verify. Send EITHER bibliography (raw BibTeX / RIS / CSL-JSON text; format auto-detected, override with format) OR claims (an array of {title + one identifier} objects) — not both. Optional options.checks defaults to ["retraction"]; pass [] to skip the retraction lookup. Max 25 entries per call (excess dropped and reported via truncated). Returns { ok, format, entries[], parseErrors[], truncated, summary } — each entry carries the same verdict / confidence semantics as /api/verify plus a retraction: { checked, doi, isRetracted, hasCorrections, hasConcern, notices } block, and summary = { total, matched, mismatch, ambiguous, not_found, errored, retracted }. One bad entry becomes status: "error" without failing the batch (a total outage returns 502). This audits citation identity — it does not check whether each source supports the claim it is cited for.

Pitfalls

• Never scrape the web UI — the JSON API is faster and stable.
• Pass identifiers verbatim; don't strip prefixes.
• Body fields differ per endpoint: format/export use text; retraction-check/oa-check use id (one identifier per call); verify wraps fields in claimed; audit takes bibliography (raw text) or claims (array), never both. Don't mix them up.
• ISBNs have no DOI, so retraction/OA return a "no DOI" result for books.
• Don't fabricate a fallback: if a call fails or returns ok:false, report that — never invent a citation, retraction status, OA verdict, or a "matched" verdict.

Verification

• curl -sS https://scholar-sidekick.com/api/health returns { "ok": true, … }.
• A good /api/format response has items[].formatted non-empty.

Optional: MCP server (power users)

Scholar Sidekick is also an MCP server (tools: resolveIdentifier, formatCitation, exportCitation, checkRetraction, checkOpenAccess, verifyCitation, auditBibliography). That path requires connecting the server in an MCP host, so the REST calls above are the zero-setup default. The MCP server works anonymously (no key required). See the companion scholar-sidekick-mcp skill, or:

npx -y scholar-sidekick-mcp@latest   # anonymous; optional SCHOLAR_API_KEY/RAPIDAPI_KEY raise limits