FAERS — PolarisRay

Connect

MCP endpoint

https://mcp.polarisray.com/faers

Transport: http · Auth: none · See the connection guide for client-specific steps.

Tools

search_reports Search adverse-event cases by drug, reaction, indication, outcome, manufacturer, country, age, sex and date range. Set suspect_only to require the drug be the reporter's suspect rather than a concomitant medication.
semantic_search_reports Find cases by meaning ("liver failure after starting a cholesterol drug") rather than exact terminology, optionally fused with keyword search.
get_report Retrieve one case in full by case number — every drug, reaction, indication and therapy period.
aggregate_reports Count cases grouped by drug, reaction, outcome, manufacturer, reporter type, country, age group, or year to see a reporting trend.
top_reactions The reactions most often co-reported with a given drug, restricted by default to cases where it was a suspect drug.
resolve Normalize free-text terms to facet labels for precise filtering.

Example questions

"What reactions are most often reported with semaglutide?"
"How has reporting for montelukast neuropsychiatric events changed since 2019?"
"Find reports of serious liver injury in patients under 18."
"Which outcomes appear most often in reports naming amiodarone as a suspect drug?"

About this dataset

FAERS is FDA’s post-marketing safety surveillance database for approved drugs and therapeutic biologics. Reports arrive from manufacturers (the large majority), health professionals and consumers, and each one records the drugs involved, the reactions observed as MedDRA Preferred Terms, why the drug was given, the outcome, and basic patient demographics.

PolarisRay loads the entire public collection — all 90 quarterly ASCII extracts from 2004Q1 to the present — into a dedicated faers database and index, separate from DailyMed / MAUDE / ClinicalTrials / Recalls.

One document per case, across the 2012 format change

FDA changed the extract format between 2012Q3 and 2012Q4, replacing the legacy LAERS ISR record key with primaryid. Both eras carry the same case-number series, so PolarisRay keys documents on the case and resolves each one to its latest version — a case first reported under LAERS and followed up years later under FAERS is one document, not two. Each case’s drugs, reactions, outcomes and therapy dates come from that same version, so nothing is double-counted across follow-ups.

What you can ask

Filter by drug (with or without requiring it to be a suspect drug rather than a concomitant one), reaction, indication, seriousness, outcome, reporter type, country, patient age and sex, and report year; retrieve a single case in full; or aggregate across the corpus to see how reporting for a drug has moved over two decades.

Source links

FDA does not publish a public per-case web page for FAERS the way it does for MAUDE. Search hits include a source_url to the quarterly extract landing page, not an individual case. Use get_report with the case number for the full record in this connector, and treat the extract page as provenance, not a primary case URL.

Reading the numbers

FAERS counts are reports, not rates. A drug that is prescribed widely, has been on the market longer, or has been in the news will accumulate more reports without being more dangerous. Every result returned by these tools carries that caveat, and top_reactions defaults to suspect-drug-only precisely because a common concomitant medication would otherwise appear attached to nearly every reaction in the database.