Build on Litmus.

Litmus runs on cleaned public data: 214,788 products and 75,113 ingredients from the NIH label database, 359 evidence summaries traced to their fact sheets, and 1,091 PubMed citations ranked by NIH iCite. The long-term bet is that this layer is infrastructure, and infrastructure is for building on.

What’s exposed

• Catalog search: products by name, brand, or ingredient.
• Evidence: sourced summaries per ingredient and outcome (no invented facts; the AI-extraction posture is documented on the provenance page).
• Papers: PubMed citations per ingredient, quality-ranked. Citations only: abstracts and full text belong to the publishers, and that constraint flows through to you.
• Starter stacks: the curated goal-to-stack recommender.

The endpoint reference → (generated from the same OpenAPI spec the router is drift-tested against)

Access

The endpoints under /public/* need no key and are IP-rate-limited; they exist so the site (and you) can read the basics. Higher limits and the keyed surface come with API keys; issuance is rolling out, write hello@nomorro.com to get in early.

The rules

• Attribution: the underlying sources are NIH DSLD/ODS, PubMed, and openFDA; keep their notices intact.
• No repackaging the feed as medical advice; the data describes evidence, it does not prescribe.
• Rate limits are enforced and generous; scraping around them is the one way to lose access.

Machine-readable summary: /llms.txt