13 New FDA Datasets: Drug Labels, Device UDI, Drug Shortages, and More

FDA Data MCP now ingests 29 openFDA datasets and exposes 33 MCP tools for AI agents. This update adds 13 new data sources and 10 new tools covering drug labels, 4.9 million device UDI records, drug shortages, device recalls, consumer and veterinary adverse events, substance identifiers, and more.

Every new dataset is normalized, deduplicated, and available through the same MCP interface you already use for recalls, 510(k) clearances, and inspections. Two of the largest datasets — drug labels and device UDI — are ingested using a streaming pipeline that processes gigabytes of data without exceeding memory limits. If your AI agent is connected to FDA Data MCP, all tools are already available.

1. What's New

This release adds 13 openFDA datasets organized into two categories: queryable datasets with dedicated MCP tools, and reference datasets that enrich existing queries.

Dataset	Records	MCP Tool
Drug Labels (SPL)	254,196	fda_drug_labels
Device UDI	4,903,764	fda_device_udi
Drug Shortages	~1,800	fda_drug_shortages
Device Recalls	~57,000	fda_device_recalls
Food Adverse Events	~220,000	fda_consumer_events
Cosmetic Adverse Events	~12,000
Veterinary Adverse Events	~1,320,000	fda_vet_events
Tobacco Problem Reports	~1,300	fda_tobacco_problems
Substance UNII	~171,000	fda_substance_lookup
NSDE (Drug Establishments)	~649,000	fda_search_nsde
Clinical Result Letters	~400	fda_clinical_result_letters
COVID-19 Serology Tests	~13,400	Reference only
Historical Documents	~8,900	Reference only

2. Drug Labels (SPL)

The largest new dataset by content volume: 254,196 structured product labels covering every FDA-approved drug. Each label includes the full prescribing information broken into searchable sections: indications, dosage, warnings, boxed warnings, contraindications, adverse reactions, drug interactions, and more.

The fda_drug_labels tool searches by drug name, manufacturer, or application number. You can request specific sections to keep responses focused.

Prompt: "What are the boxed warnings for metformin?"
Tool:   fda_drug_labels { drug_name: "metformin", section: "boxed_warning" }

3. Device UDI

Nearly 5 million Unique Device Identification records covering every medical device on the US market. Each record includes brand name, manufacturer, device description, product codes, GMDN terms, sterilization details, and premarket submission references.

The fda_device_udi tool searches by brand name, company, product code, or DI number.

Prompt: "Find all cardiac monitors made by Medtronic"
Tool:   fda_device_udi { company_name: "Medtronic", product_code: "MHX" }

4. Drug Shortages

The FDA tracks current and resolved drug shortages — which drugs are unavailable, why, and which companies are affected. This data is critical for supply chain planning, formulary management, and understanding market dynamics.

The fda_drug_shortages tool lets you search by generic name, status (currently in shortage, resolved, discontinued), or company name. Each record includes the generic name, dosage form, company, status, and availability details.

Prompt: "Which injectable drugs are currently in shortage?"
Tool:   fda_drug_shortages { status: "Currently in Shortage" }

We also launched a Drug Shortage Statistics dashboard with live charts showing shortage status breakdowns, top affected companies, and dosage form analysis.

5. Device Recalls

While our existing fda_search_recalls tool covers enforcement-level recall data (from the FDA's enforcement reports), the new fda_device_recalls tool pulls from the device-specific recall database. This source includes recall event numbers, product codes, and device-specific metadata not available in enforcement reports.

Prompt: "Find device recalls for product code LLZ in the past year"
Tool:   fda_device_recalls { product_code: "LLZ", date_from: "2025-02-06" }

6. Consumer Adverse Events

Food and cosmetic adverse event reports are now available through a single fda_consumer_events tool. These reports cover consumer complaints about food products (contamination, allergic reactions, illness) and cosmetics (skin reactions, allergic responses).

The tool accepts a product_area filter (food or cosmetic) along with product name, reaction, and date filters.

Prompt: "Show me food adverse events involving peanut allergies from 2025"
Tool:   fda_consumer_events { product_area: "food", reaction: "peanut", date_from: "2025-01-01" }

7. Veterinary Adverse Events

With over 1.3 million reports, the veterinary adverse event database is the largest new dataset in this release. It tracks adverse reactions to animal drugs and contains species, breed, drug, and outcome information.

Prompt: "What are the most reported adverse events for dogs taking flea medication?"
Tool:   fda_vet_events { species: "Dog", drug_name: "flea" }

8. Reference Datasets

Substance UNII Lookup

The UNII (Unique Ingredient Identifier) database maps substance names to their standardized identifiers. Use fda_substance_lookup to resolve ingredient names or look up UNII codes.

Prompt: "What is the UNII code for acetaminophen?"
Tool:   fda_substance_lookup { substance_name: "acetaminophen" }

NSDE (National Drug Code Directory)

The NSDE maps NDC codes to drug names, dosage forms, and application numbers. Use fda_search_nsde to look up drugs by proprietary name, NDC, or application number.

Clinical Result Letters

Complete Response Letters (CRLs) and Approval letters for drug applications. Search by company name, application number, or letter type using fda_clinical_result_letters.

Tobacco Problem Reports

Consumer-reported problems with tobacco products, searchable by product type and health problem via fda_tobacco_problems.

COVID-19 Serology & Historical Documents

These two smaller datasets are ingested as reference tables. COVID-19 serology test performance data (~13,400 records) and FDA historical documents (~8,900 records) are available in the database but do not have dedicated MCP tools.

9. Example Queries

Here are practical queries you can run with the new tools:

• Drug labeling: "What are the boxed warnings and contraindications for warfarin?" — pull specific SPL sections from 254K structured drug labels.
• Device identification: "Find all single-use Class II devices made by [Company]" — search 4.9M UDI records by manufacturer and attributes.
• Supply chain: "Are any of our active ingredients currently in drug shortage?" — cross-reference your drug formulations with the shortage database.
• Competitive intelligence: "How many device recalls has [Company] had in the past 3 years, broken down by product code?"
• Safety signal detection: "Show me consumer adverse events mentioning [ingredient] in the past 6 months" — identify emerging safety signals in food or cosmetic products.
• Veterinary pharmacovigilance: "What drugs have the most adverse events reported in cats?" — filter by species and aggregate by drug.
• Substance resolution: "Look up the UNII for ibuprofen and find all NSDE entries for that ingredient."

10. What's Next

With 29 datasets and 33 tools now live, we're evaluating the feasibility of ingesting the two largest remaining openFDA datasets:

• FAERS (Drug Adverse Events): ~20 million reports covering post-market drug safety signals, patient demographics, and drug-reaction associations.
• MAUDE (Device Adverse Events): ~23.6 million medical device reports with event narratives, device details, and patient outcomes.

These datasets require a database plan upgrade and quarterly incremental ingestion. Adding them would make FDA Data MCP the most comprehensive FDA data source available through MCP, covering over 50 million records across 31 datasets.

Further Reading

• FDA Data MCP Documentation — full tool reference for all 33 tools
• Drug Shortage Statistics Dashboard — live charts and data
• Recall Statistics Dashboard — enforcement trends
• Using AI to Analyze FDA Compliance Data — setup guide
• How FDA Data MCP Works — technical deep dive