Academic Research Skill
Free, citation-grounded academic / scientific / medical literature research using verified free APIs + our existing model arsenal — no paid SaaS subscription required. Built as honest, working equivalents of popular research SaaS products.
Created: 2026-05-31 · Status: ✅ tested with live data
---
🔴 HONEST API STATUS — the "research SaaS" tools (verified 2026-05-31)
د. وائل asked to "download" Elicit / Consensus / OpenEvidence / Scite. The truth, verified by checking each vendor's actual docs/pricing pages:
| Tool | Public API? | Free? | Reality |
|---|---|---|---|
| Elicit (elicit.com) | ✅ Yes — docs.elicit.com | ❌ No | API is in preview for Pro/Teams paid users only. Searches 138M+ papers + generates research reports. Requires paid subscription. |
| Consensus (consensus.app) | ✅ Yes — consensus.app/home/api | ❌ No | Credential-based paid API. Returns relevant papers + citations. Requires subscription/credits. |
| Scite (scite.ai) | ✅ Yes — api.scite.ai/docs + MCP | ❌ No | Most endpoints need a token (contact sales@scite.ai); credit-based paid tiered pricing. Only tallies/papers endpoints are token-free. Built on licensed 40+ publisher corpus. |
| OpenEvidence (openevidence.com) | ❌ No public dev API | n/a | Clinician-only web product (verified U.S. clinicians). NEJM/JAMA official AI partner. No programmatic access for third parties. |
Conclusion: None of these can be "downloaded" or used free via API. They are commercial web products. So this skill rebuilds their core value from free APIs.
---
✅ WHAT WE BUILT (free, tested, working)
All scripts in scripts/. No API key needed for the core scholarly sources.
Free scholarly APIs used (all verified live 2026-05-31)
| API | Key needed? | Status | Notes |
|---|---|---|---|
| OpenAlex (api.openalex.org) | ❌ None | ✅ Primary | 250M+ works, fully free, fast, abstracts (inverted index), citations. Best free source. |
| Europe PMC (ebi.ac.uk/europepmc) | ❌ None | ✅ Works | Biomed-heavy, full-text, abstracts, MED+PMC. |
| Crossref (api.crossref.org) | ❌ None | ✅ Works | DOI metadata, broad coverage. |
| arXiv (export.arxiv.org) | ❌ None | ✅ Works (use https) | Preprints (CS/physics/math). |
| Semantic Scholar (api.semanticscholar.org) | optional | ⚠️ Rate-limited | Public pool returns 429 (Too Many Requests) often. Scripts back off + degrade gracefully. Set SEMANTIC_SCHOLAR_API_KEY (free, request at semanticscholar.org/product/api) for reliable use. |
| PubMed/NCBI E-utilities | optional | ✅ Works | Via existing medical-arsenal/scripts/pubmed_search.py. Set NCBI_API_KEY for higher rate. |
---
Scripts
1. lit_search.py — free equivalent of Elicit's literature search
Aggregates + de-duplicates (by DOI/title) + ranks (citations, then year)
results across OpenAlex + Europe PMC + Crossref + arXiv + (optional) Semantic Scholar.bash
python3 scripts/lit_search.py "CRISPR gene editing safety" --limit 10 --text
python3 scripts/lit_search.py "long covid fatigue" --sources openalex,europepmc --year-from 2020
python3 scripts/lit_search.py "diffusion models" --sources arxiv --limit 5
--limit --sources (csv) --year-from --per-source --text (else JSON).
Sources: openalex,semanticscholar,europepmc,crossref,arxiv.
Default sources: openalex,europepmc,crossref (the most reliable free trio).2. consensus_check.py — free equivalent of Consensus.app's consensus meter
Takes a scientific claim, pulls the most relevant/cited papers (OpenAlex +
Europe PMC), then asks an arsenal LLM to judge each paper SUPPORT / CONTRADICT /
MIXED / IRRELEVANT and produce an overall consensus % + verdict + summary.bash
python3 scripts/consensus_check.py "Intermittent fasting improves insulin sensitivity" --text
python3 scripts/consensus_check.py "Vitamin D prevents respiratory infections" --limit 12 --year-from 2015
python3 scripts/consensus_check.py "..." --model claude --use-s2
--limit --year-from --model gemini|claude --use-s2 --text.
Summarizer default Gemini 2.5 Pro → Claude Opus 4.7 fallback. Needs
GOOGLE_API_KEY/GEMINI_API_KEY or ANTHROPIC_API_KEY (in env.sh).3. medical_evidence.py — free equivalent of OpenEvidence
For clinical questions: merges PubMed (via medical-arsenal) + Europe PMC,
prioritizes high-evidence study types, and (optionally) synthesizes a
citation-grounded evidence summary with an arsenal LLM. Always returns
citations + a safety disclaimer.bash
python3 scripts/medical_evidence.py "metformin for PCOS insulin resistance" --synthesize --text
python3 scripts/medical_evidence.py "statin primary prevention elderly" --filter meta-analysis --years 5
--limit --years --filter (meta-analysis|systematic-review|rct|guideline|review)
--synthesize --model gemini|claude|auto --text.
SAFETY: research aid only — no diagnosis/prescribing. Defer to a licensed
clinician. Emergencies (Kuwait): 112. Pairs with skills/medical-arsenal/.---
When to use which
- Literature review / "what does research say on X" / find papers →
lit_search.py - "Is claim X true? what's the scientific consensus?" →
consensus_check.py - Clinical / medical / drug / disease evidence question →
medical_evidence.py
skills/medical-arsenal/ for drug lookup, lab interpretation)If د. وائل wants the real SaaS products (paid)
Report honestly: Elicit (Pro/Teams), Consensus (subscription), Scite (contact sales, credit-based) all need paid subscriptions for API access. OpenEvidence has no public API at all (clinician-only). Our free stack covers ~90% of their practical value (structured search, consensus judgment, evidence synthesis with citations) at $0 using OpenAlex/Europe PMC/PubMed/Crossref/arXiv + our LLMs. The main thing the paid tools add: proprietary licensed full-text corpora (Scite's 40+ publisher deals, OpenEvidence's NEJM/JAMA partnership) and polished UIs.Dependencies
- Python 3 (stdlib only —
urllib,json,xml; no pip installs needed) - env.sh keys:
GOOGLE_API_KEY/GEMINI_API_KEY,ANTHROPIC_API_KEY(for synthesis) - Optional:
SEMANTIC_SCHOLAR_API_KEY,NCBI_API_KEY(higher rate limits) - Sibling skill:
skills/medical-arsenal/scripts/pubmed_search.py