Connectors

SaberTools researches across 18 live public-API connectors. Each project has a domain, and the platform's domain routing automatically selects the right connectors for that domain — so a health query hits USDA, PubMed, and ClinicalTrials.gov, while a bibliographic query hits OpenAlex, Crossref, and DOAJ.

The catalogue

API	Domain	Auth
USDA FoodData Central	health	Free key / DEMO_KEY
NIH DSLD	health	None
PubMed E-utilities	health / bibliographic	None
OpenAlex	bibliographic	Free key (optional)
Crossref	bibliographic	None
Semantic Scholar	scientific	None
arXiv	scientific	None
ClinicalTrials.gov v2	health	None
openFDA drug labels	health	Free key (optional)
World Bank Indicators	government	None
Data.gov Catalog	government	DEMO_KEY
OpenAlex abstract reconstruct	bibliographic	None
Jina Reader (scraper)	open_web	None
Google Scholar (via Claude)	bibliographic	None (Claude-proxied)
Wikipedia API	open_web	None
Open Library	cultural_heritage	None
FRED (Federal Reserve)	finance	Free key
DOAJ	bibliographic	None

Domain routing

A project's domain drives which connectors run. The routing table is extensible — domains map to an ordered set of connectors, and the agent queries them in parallel, deduplicating and grading results as they arrive.

Domain	Typical connectors
health	USDA, NIH DSLD, PubMed, ClinicalTrials.gov, openFDA
bibliographic	OpenAlex, Crossref, DOAJ, Google Scholar
scientific	Semantic Scholar, arXiv, PubMed
government	World Bank, Data.gov
finance	FRED
open_web	Wikipedia, Jina Reader
cultural_heritage	Open Library

Credibility & grading

Every connector's results are normalized into the same source shape — title, type, DOI, authors, evidence grade, clinical status, credibility score (1–10), and license (license_label + rights_url). That uniformity is what makes the output a valid UKS packet regardless of which APIs produced it.

→ Related: Evidence grading · Export to UKS · API reference