openFDA Scraper · Drug Events, Recalls, Labels & Devices

Scrape openFDA - the FDA's public API - for drug adverse events, drug and food recalls, product labels, the NDC directory, medical device events, device recalls and 510(k) clearances. No API key required, no login, no headless browser. Run it empty and you get the newest drug recalls; pick a different dataset and it returns that instead.

Maintained by reapX. Every row cites the Apify run that produced it — nothing is inferred, modelled or filled in, and a field absent from the source is absent from the row. The extracted archive for this source is browsable at reapx.dev/data/openfda-scraper/ and mirrored as an open dataset on Hugging Face and Kaggle. Questions: reapxdev@proton.me

Eight FDA datasets, over 46 million records, every row keyed by its own FDA identifier.

What this openFDA scraper returns

Dataset	Records	Keyed by
Drug recalls and enforcement	17,800+	Recall number
Drug adverse events (FAERS)	20.3M	Safety report ID
Drug product labels (SPL)	261,000+	SPL set ID
NDC directory of marketed drugs	136,000+	Product NDC
Device recalls	58,800+	Product recall number
Device adverse events (MAUDE)	25.3M	Report number
510(k) device clearances	175,000+	K number
Food recalls and enforcement	29,200+	Recall number

Every row carries recordId, recordType, endpoint, primaryDate and scrapedAt, plus the fields that actually apply to that dataset - recall class and reason, reaction terms and seriousness flags, label sections, ingredients and routes, applicant and review time.

Five named dataset views ship with it: Records, Recalls, Adverse events, Products & labels and 510(k) clearances.

How to scrape openFDA

1. Pick a dataset. Leave everything else empty and you get the newest records.
2. Narrow by product or brand name, company, recall class, status, state or a date window.
3. For anything the plain-language filters do not cover, use the raw openFDA query field.
4. Set Max records. You are charged once per record returned, so that is your cost cap.

What you can build with it

• Recall monitoring - run daily on drug, device or food recalls and diff by recordId.
• Pharmacovigilance signals - count reaction terms for a drug across FAERS over a date window.
• Competitive device intelligence - every 510(k) an applicant has cleared, with review times.
• Label change tracking - pull SPL labels by set ID and watch indications and warnings move.
• Formulary and NDC datasets - the marketed drug directory with ingredients, routes and DEA schedule.
• Food safety alerting - Class I food recalls by state, as they are published.
• Regulatory RAG corpora - label text and recall reasons with a citable FDA identifier on every row.

Input configuration

Every field is optional. The prefilled values below are what runs if you press Start without changing anything.

What to read

Leave every filter below empty and the run returns the newest records in the chosen dataset.

Field	Type	Accepts	What it does
dataset	string	one of 8: drug-recalls, drug-adverse-events, drug-labels, drug-ndc-directory, ...	Which openFDA endpoint to read. Each one returns a different kind of record, all keyed by their own FDA identifier. Prefilled with "drug-recalls".
searchTerm	string	free text	Matches the product, brand or device name fields of the chosen dataset.
firmName	string	free text	Recalling firm, labeler, applicant or manufacturer, depending on the dataset.

How much to fetch

Volume and ordering. You pay per record returned, so this decides the price of the run.

Field	Type	Accepts	What it does
maxRecords	integer	1 to 25000	Hard cap on how many records the run returns. This is also your cost cap: you are charged once per record returned. openFDA serves at most 25,000 records for one query. Prefilled with 100.
sortBy	string	newest, oldest, none	Ordering on the dataset's main date field. Prefilled with "newest".

Dates

All dates are YYYY-MM-DD and apply to whichever date field the chosen dataset uses.

Field	Type	Accepts	What it does
lastNDays	integer	1 to 20000	Rolling window on the dataset's main date - recall initiation, report received, decision date or marketing start. Ignored if you set an explicit "From date". Note the adverse event datasets lag: FAERS is usually 3-5 months behind and MAUDE 1-2 months, so windows under 180 days can come back empty.
dateFrom	string	free text	Earliest date, YYYY-MM-DD.
dateTo	string	free text	Latest date, YYYY-MM-DD.

Recall filters

These apply to the recall and enforcement datasets.

Field	Type	Accepts	What it does
classification	string	Class I, Class II, Class III	Class I is the most serious (reasonable probability of serious harm), Class III the least. Applies to drug and food recalls.
recallStatus	string	one of 5: Ongoing, Completed, Terminated, Pending, ...	Where the recall stands today. Drug and food recalls use Ongoing/Completed/Terminated; device recalls use Open, Classified/Terminated/Completed.
state	string	free text	Two-letter state code of the recalling firm or applicant.
country	string	free text	Country of the recalling firm or applicant.

Adverse event filters

These apply to the drug adverse event dataset.

Field	Type	Accepts	What it does
seriousOnly	true/false	checkbox	Keep only adverse event reports FDA flagged as serious. Applies to drug adverse events.
deathsOnly	true/false	checkbox	Keep only adverse event reports where death was an outcome.
reaction	string	free text	MedDRA reaction term as FDA records it, in capitals, e.g. HEADACHE, NAUSEA, ANAPHYLACTIC REACTION.

Advanced

Escape hatches for anything the plain-language filters above do not cover.

Field	Type	Accepts	What it does
searchQuery	string	free text	Advanced. A raw openFDA search expression, combined with every other filter using AND. Example: openfda.pharm_class_epc:"Antibacterial". Full syntax at open.fda.gov/apis/query-syntax.
apiKey	string	free text	Optional and free from open.fda.gov/apis/authentication. Without one, openFDA allows 1,000 requests per day per IP; with one, 120,000. Only needed for very large runs.

Pricing

Pay per event. The primary event is Record returned (record-returned), charged at $0.002 per event on the free plan, falling to $0.0008 on the highest tier.

Charged once per complete openFDA record written to the default dataset. A rate-limited or unreachable API, a rejected query and a search that matches nothing are never charged.

Minimum charge cap per run: $0.002.

Tiered discounts apply automatically on every paid Apify plan. Platform usage is absorbed by the Actor, so the per-event price is the whole price.

Usage examples

The prefilled run, which is what Start does with nothing changed

{
  "dataset": "drug-recalls",
  "maxRecords": 100,
  "sortBy": "newest"
}

Filter by product or brand name

{
  "dataset": "drug-recalls",
  "maxRecords": 100,
  "sortBy": "newest",
  "searchTerm": "insulin"
}

Filter by company or firm

{
  "dataset": "drug-recalls",
  "maxRecords": 100,
  "sortBy": "newest",
  "firmName": "Medtronic"
}

Filter by in the last n days

{
  "dataset": "drug-recalls",
  "maxRecords": 100,
  "sortBy": "newest",
  "lastNDays": 90
}

Filter by from date

{
  "dataset": "drug-recalls",
  "maxRecords": 100,
  "sortBy": "newest",
  "dateFrom": "2026-01-01"
}

Output example

One row, exactly as the actor wrote it to the dataset:

{
  "recordId": "65585-361",
  "recordType": "drug-ndc",
  "endpoint": "drug/ndc",
  "primaryDate": "2026-07-31",
  "scrapedAt": "2026-08-03T01:57:36Z",
  "brandName": "WhiskCare 373",
  "genericName": "Alcohol",
  "labelerName": "Whisk Products, Inc.",
  "dosageForm": "SOLUTION",
  "routes": [
    "TOPICAL"
  ],
  "marketingCategory": "OTC MONOGRAPH DRUG",
  "productType": "HUMAN OTC DRUG",
  "applicationNumber": "505G(a)(3)",
  "activeIngredients": [
    "ALCOHOL 700 mL/L"
  ],
  "listingExpirationDate": "2027-12-31",
  "packageCount": 7
}

Fields on every row

Field	Type	What it is
activeIngredient	string	Active ingredient text from the label. Present on: drug-labels.
activeIngredients	list	Active ingredients with their strengths. Present on: drug-ndc-directory.
adverseEventFlag	string	Whether the report was flagged as an adverse event. Present on: device-adverse-events.
advisoryCommittee	string	FDA advisory committee that reviews this device area. Present on: device-510k-clearances.
applicant	string	Company that submitted the 510(k). Present on: device-510k-clearances.
applicationNumber	string	FDA application number, e.g. an NDA or ANDA. Present on: drug-ndc-directory.
brandName	string	Marketed brand name of the drug product. Present on: drug-ndc-directory.
brandNames	list	Brand names FDA associates with the product. Present on: drug-labels, drug-recalls.
cfresId	string	Internal FDA device recall record ID. Present on: device-recalls.
city	string	City of the firm on record. Present on: device-510k-clearances, device-recalls, drug-recalls, food-recalls.
classification	string	FDA seriousness class. Class I means a reasonable probability of serious harm. Present on: drug-recalls, food-recalls.
clearanceType	string	Type of 510(k) submission. Present on: device-510k-clearances.
companyNumber	string	Reporting company case number. Present on: drug-adverse-events.
country	string	Country of the firm on record. Present on: device-510k-clearances, drug-recalls, food-recalls.
dateOfEvent	string	Date the device event occurred, YYYY-MM-DD. Present on: device-adverse-events.
datePosted	string	Date the device recall was posted, YYYY-MM-DD. Present on: device-recalls.
dateReceived	string	Date FDA received the submission, YYYY-MM-DD. Present on: device-510k-clearances.
deaSchedule	string	Controlled substance schedule, when the drug is scheduled. Present on: drug-ndc-directory.
decisionCode	string	Short FDA code for the decision. Present on: device-510k-clearances.
decisionDescription	string	FDA decision on the submission. Present on: device-510k-clearances.
deviceBrandNames	list	Brand names of the devices in the report. Present on: device-adverse-events.
deviceCount	integer	How many devices this report covers. Present on: device-adverse-events.
deviceGenericNames	list	Generic names of the devices in the report. Present on: device-adverse-events.
deviceName	string	Trade name of the cleared device. Present on: device-510k-clearances.
distributionPattern	string	Where the product was distributed. Present on: device-recalls, drug-recalls, food-recalls.
documentId	string	Version-specific SPL document ID. The set ID in recordId is stable across versions. Present on: drug-labels.
dosageAndAdministration	string	Dosing instructions from the label. Present on: drug-labels.
dosageForm	string	Physical form of the product, e.g. TABLET. Present on: drug-ndc-directory.
drugCount	integer	How many drugs this report lists. Present on: drug-adverse-events.
endpoint	string	The api.fda.gov endpoint the row came from.
eventId	string	FDA enforcement event ID grouping related recall records. Present on: drug-recalls, food-recalls.
eventType	string	Nature of the device event, e.g. Malfunction or Injury. Present on: device-adverse-events.
genericName	string	Generic name of the drug product. Present on: drug-ndc-directory.
genericNames	list	Generic drug names FDA associates with the product. Present on: drug-labels, drug-recalls.
indicationsAndUsage	string	What the product is approved to treat, from the label. Present on: drug-labels.
initialFirmNotification	string	How the firm first told its customers. Present on: drug-recalls, food-recalls.
kNumbers	list	510(k) clearance numbers linked to the recalled device. Present on: device-recalls.
labelVersion	string	Version number of this SPL label. Present on: drug-labels.
labelerName	string	Company that labels and markets the product. Present on: drug-ndc-directory.
listingExpirationDate	string	Date the NDC listing expires, YYYY-MM-DD. Present on: drug-ndc-directory.
manufacturerNames	list	Manufacturer names on the record. Present on: device-adverse-events, drug-labels.
marketingCategory	string	Regulatory pathway under which the product is marketed. Present on: drug-ndc-directory.
medicinalProducts	list	Drug products named in the report. Present on: drug-adverse-events.
occurCountry	string	Country where the event occurred. Present on: drug-adverse-events.
packageCount	integer	How many package configurations this NDC has. Present on: drug-ndc-directory.
patientOnsetAge	string	Patient age when the reaction began, as reported. Present on: drug-adverse-events.
patientSex	string	Patient sex as reported. Present on: drug-adverse-events.
pharmClasses	list	Established pharmacologic class terms. Present on: drug-ndc-directory.
primaryDate	string	The dataset’s main date, normalised to YYYY-MM-DD: recall initiation, report received, decision date or marketing start.
productCode	string	FDA three-letter device product code. Present on: device-510k-clearances, device-recalls.
productCodes	list	FDA device product codes in the report. Present on: device-adverse-events.
productDescription	string	Full description of the recalled or reported product as the firm filed it. Present on: device-recalls, drug-recalls, food-recalls.
productNdcs	list	National Drug Codes covered by this label. Present on: drug-labels.
productQuantity	string	How much product is covered by the recall, as filed. Present on: device-recalls, drug-recalls, food-recalls.
productType	string	Broad FDA product category. Present on: drug-ndc-directory, drug-recalls, food-recalls.
purpose	string	Purpose section of the label. Present on: drug-labels.
reactionCount	integer	How many distinct reactions this report lists. Present on: drug-adverse-events.
reactions	list	MedDRA reaction terms reported. Present on: drug-adverse-events.
reasonForRecall	string	Why the product was recalled, in the firm’s own words. Present on: device-recalls, drug-recalls, food-recalls.
recallAction	string	What the firm is doing about the recalled device. Present on: device-recalls.
recallStatus	string	Where the recall stands today. Present on: device-recalls, drug-recalls, food-recalls.
recallingFirm	string	Company that initiated or was ordered to conduct the recall. Present on: device-recalls, drug-recalls, food-recalls.
receiptDate	string	Date FDA most recently received information, YYYY-MM-DD. Present on: drug-adverse-events.
recordId	string	The FDA identifier for this record - recall number, NDC, 510(k) number, safety report ID or SPL set ID depending on the dataset. Stable, unique, and how the row is addressed downstream.
recordType	string	Which kind of FDA record this row is.
reportDate	string	Date FDA published the enforcement report, YYYY-MM-DD. Present on: drug-recalls, food-recalls.
reportSource	string	Who submitted the report. Present on: device-adverse-events.
reviewDays	integer	Days between FDA receiving the 510(k) and deciding it. Present on: device-510k-clearances.
routes	list	Routes of administration. Present on: drug-labels, drug-ndc-directory.
scrapedAt	string	UTC ISO 8601 timestamp of this run.
serious	true/false	FDA flagged the report as serious. Present on: drug-adverse-events.
seriousnessDeath	true/false	Death was an outcome of this report. Present on: drug-adverse-events.
seriousnessHospitalization	true/false	Hospitalisation was an outcome of this report. Present on: drug-adverse-events.
seriousnessLifeThreatening	true/false	The event was life threatening. Present on: drug-adverse-events.
state	string	Two-letter US state code of the firm on record. Present on: device-510k-clearances, device-recalls, drug-recalls, food-recalls.
substanceNames	list	UNII substance names on the label. Present on: drug-labels.
terminationDate	string	Date FDA terminated the recall, YYYY-MM-DD. Present on: food-recalls.
voluntaryMandated	string	Whether the firm initiated the recall or FDA required it. Present on: drug-recalls, food-recalls.
warnings	string	Warnings section of the product label. Present on: drug-labels.

5 named dataset views ship with it: Records, Recalls, Adverse events, Products & labels, 510(k) clearances. They drive the Output tab in Console and the Output block on the Actor's .md page.

FAQ

How do I scrape openFDA data?

Pick a dataset and run. Everything comes from api.fda.gov, the FDA's own public API, which needs no authentication for normal use.

Do I need an openFDA API key?

Not for most runs. Without a key openFDA allows 1,000 requests per day per IP; this actor pulls 1,000 records per request, so that is a great deal of data. For very large jobs, paste a free key from open.fda.gov/apis/authentication into the optional openFDA API key field and the limit rises to 120,000 requests per day.

How do I find Class I recalls?

Choose Drug recalls or Food recalls, set Recall classification to Class I, and optionally add a date window. Class I means FDA judged there is a reasonable probability of serious harm or death.

How do I search adverse events for one drug?

Choose Drug adverse events (FAERS) and put the drug name in Product or brand name. Add a Reaction term such as ANAPHYLACTIC REACTION to narrow to one reaction, or turn on Reports involving death only.

How far back does the data go?

It varies by dataset - FAERS runs from 2004, enforcement reports from 2012, 510(k) clearances from 1976. Set From date as early as you like; the actor passes it straight through.

How many records can one run return?

Up to 25,000 for a single query, which is openFDA's own paging ceiling. Split a larger job into date windows and run each one.

What identifies each row?

recordId - the recall number, safety report ID, SPL set ID, NDC, K number or device recall number, depending on the dataset. It is FDA's own identifier and is stable across runs.

Can I use my own openFDA query syntax?

Yes. The Raw openFDA query field takes any expression the API accepts, for example openfda.pharm_class_epc:"Antibacterial", and it is combined with the other filters using AND.

Can an AI agent call this actor?

Yes. Every input and output field carries a title, description and example, and every row is addressed by its FDA recordId, so an agent can fill the input and chain the output without extra prompting.

---

Unofficial - not affiliated with the FDA. Collects public data only. reapx. Contact reapxdev@proton.me.

The full published archive

• openfda scraper archive — every entity this Actor has observed
• All reapx datasets — 20 sources, tens of thousands of pages
• llms.txt · feed.json · feed.xml — the machine-readable index, for agents

Each archive page carries Dataset JSON-LD, a canonical URL and the identifiers used, so an agent can resolve an entity without running anything. Nothing on those pages is estimated or modelled.

🧪 Example input

A real, runnable configuration — this is an actual input this Actor has run with.

{
  "dataset": "drug-ndc-directory",
  "lastNDays": 365,
  "maxRecords": 500
}

📄 Sample output

One real row from a real run of this Actor, unedited.

{
  "recordId": "65585-361",
  "recordType": "drug-ndc",
  "endpoint": "drug/ndc",
  "primaryDate": "2026-07-31",
  "scrapedAt": "2026-08-03T01:57:36Z",
  "brandName": "WhiskCare 373",
  "genericName": "Alcohol",
  "labelerName": "Whisk Products, Inc.",
  "dosageForm": "SOLUTION",
  "routes": [
    "TOPICAL"
  ],
  "marketingCategory": "OTC MONOGRAPH DRUG",
  "productType": "HUMAN OTC DRUG",
  "applicationNumber": "505G(a)(3)",
  "activeIngredients": [
    "ALCOHOL 700 mL/L"
  ],
  "listingExpirationDate": "2027-12-31",
  "packageCount": 7
}

How it works

1. You set the filters below, or none at all.
2. The Actor calls the source's own public endpoint and pages through the results.
3. Every row is pushed to the dataset as it is built, not buffered to the end — so a run that hits its time limit still returns everything it collected up to that point.
4. You are charged per row returned, so the maximum-results field is also your cost cap.

💬 Your feedback

Found a bug, or need a field this does not return yet? Open an issue on the Actor's Issues tab, or write to reapxdev@proton.me. Bugs get fixed and reasonable field requests get added.