Federated Search
LiveFederated search across MAUDE, FAERS, ClinicalTrials.gov, DailyMed, openFDA SPL, FDA Recalls, ChEMBL, and PubChem
One MCP surface over eight live connectors at once: MAUDE device adverse events, FAERS drug adverse events, ClinicalTrials.gov studies, DailyMed labels, openFDA SPL labels, FDA recalls, ChEMBL bioactivity, and PubChem compounds. Because the datasets share a normalized medical vocabulary, a single query can search, filter, and correlate across them — by keyword or by meaning. Each result is tagged with its source and includes an official `url` when available (FAERS links to its quarterly extract rather than a per-case page).
Connect
https://mcp.polarisray.com/federated
Transport: http · Auth:
none · See the
connection guide for client-specific steps.
Tools
-
federated_searchFree-text + facet search across MAUDE, FAERS, ClinicalTrials.gov, DailyMed, openFDA SPL, FDA recalls, ChEMBL, and PubChem in one call; hits tagged by source, with per-source totals and official urls. -
semantic_searchSearch federated sources by meaning, not just exact keywords — finds relevant reports, trials, labels, recalls, and compounds by concept; reports which sources were searched. -
correlateCross-source view of one concept — how many adverse-event reports vs registered trials mention it, plus the top co-occurring device types and conditions in each. -
resolveNormalize lay terms, synonyms, and abbreviations to the canonical concept labels the datasets share, for precise filtering.
Example questions
- "Search adverse events and clinical trials about sepsis together."
- "Is Medtronic seeing both adverse-event reports and active research for its devices?"
- "Find reports, trials, labels, and recalls related to insulin-pump dosing errors."
- "Have there been Class I infusion-pump recalls, and how do they compare to MAUDE volume?"
- "For 'infusion pump', compare the volume of FDA adverse-event reports vs registered trials."
About this connector
Federated Search unifies eight live connectors behind one MCP endpoint:
MAUDE, FAERS,
ClinicalTrials.gov, DailyMed,
openFDA SPL, FDA Recalls,
ChEMBL, and PubChem. (PDB
protein structures are a standalone server, not part of federated.) The datasets
share the same normalized medical
vocabulary (diseases, device types, substances, anatomy, manufacturers), so a
single query works across them and every hit is tagged with its source — plus an
official url (FDA MAUDE detail, ClinicalTrials.gov study, DailyMed / openFDA
label, or iRES Event page)
when one can be built.
FAERS drug-event cases are included in this federated endpoint too; FAERS hits link to the quarterly extract rather than a per-case page. FAERS is also available as its own server if you want only those reports.
What you can ask
Ask one plain-English question and get results spanning device and drug adverse events, clinical trials, DailyMed and openFDA drug labels, and recalls: search everything together by keyword or by meaning, or correlate a concept to see, side by side, how much post-market adverse-event activity vs active research a device or condition is seeing.
Search by meaning
semantic_search finds conceptually relevant records across federated sources by
meaning — even when the wording differs — and reports which sources were
searched.
Caveats
Federated results carry the limitations of the underlying datasets: MAUDE and
FAERS are unverified passive-surveillance data, ClinicalTrials.gov is self-reported and not
endorsed, DailyMed and openFDA labels vary in completeness, and FDA recall
classifications and status can change. Treat counts as signals, confirm against the
primary records via the returned url, and never treat the data as medical advice.