Model Context Protocol

Public data, connected to your assistant.

PolarisRay turns hard-to-search public datasets into MCP servers your AI assistant can query in plain language — FDA MAUDE, ClinicalTrials.gov, DailyMed, FDA recalls, and federated search across them.

Built for healthcare & life-sciences teams — research, safety, and regulatory.

Base MCP endpoint
https://mcp.polarisray.com/maude

Data sources

7 live, more on the way

See all →
🩺 Live

MAUDE

Manufacturer and User Facility Device Experience

Query FDA medical-device adverse event reports — by keyword or by meaning — in natural language.

medical 1991–present, ~25M+ reports
🧪 Live

ClinicalTrials.gov

ClinicalTrials.gov study registry (via CTTI AACT)

Search and analyze the world's clinical trial registry — by keyword or by meaning — in natural language.

medical 2000–present, ~500K+ studies
💊 Live

DailyMed

DailyMed FDA Structured Product Labeling (SPL)

Search and analyze official FDA drug labels — by keyword or by meaning — in natural language.

medical 160K+ current labels (Rx, OTC, homeopathic, animal)
⚠️ Live

FDA Recalls

FDA Enforcement Reports (Recalls)

Search FDA recall and enforcement actions across drugs, devices, food, and biologics.

regulatory 2012–present (iRES weekly archive)
🔗 Live

Federated Search

Federated search across MAUDE, ClinicalTrials.gov, DailyMed, and FDA Recalls

Search adverse events, clinical trials, drug labels, and recalls together — by keyword or by meaning — in one call.

medical ~25M+ MAUDE reports · ~500K+ trials · 160K+ labels · FDA enforcement recalls
🏷️ Live

openFDA SPL

openFDA Structured Product Labeling (drug labels)

Search FDA drug labels from the openFDA JSON dump — by keyword or by meaning.

medical ~262K current labels (all /drug/label partitions)
💊 Live

FAERS

FDA Adverse Event Reporting System

Search 20+ years of reported adverse drug events — drugs, reactions, outcomes, and who reported them.

regulatory 2004Q1–present (90 quarterly extracts; LAERS 2004–2012Q3, FAERS 2012Q4–)

Made for agents

Let your agent do the comparing

Point your AI agent at PolarisRay and ask it to search or compare records. Because it queries the data directly over MCP — structured results, not scraped web pages — the task is easier for the model, the answers are more accurate, and it burns far fewer tokens than reading through pages of results. Agents tend to prefer it.

Compare in one ask

"Compare device A vs device B" becomes a single structured query — not dozens of page loads to read and summarize.

Fewer tokens

No HTML to wade through and no search pages to parse, so answers come back faster and cost less.

Straight from the source

Hits include an official url to the primary record (DailyMed, ClinicalTrials.gov, FDA MAUDE detail, or iRES recalls) so you can verify what the agent found.

How it works

Three steps

1

Add the endpoint

Point your MCP-capable client (Claude, Cursor, VS Code…) at a PolarisRay endpoint. No account, no API key for public sources.

2

Ask in plain language

Your assistant discovers the available tools and translates your questions into the right queries.

3

Get sourced answers

Results come back structured, with an official url on each hit so you can open the primary record and verify everything.

Who it's for

Made for healthcare & life-sciences teams

Why PolarisRay

Fewer tokens. Search by meaning.

Answer-ready results — about 26× less context than raw APIs — plus concept search the official FDA API can't do.

See why PolarisRay →