openFDA Scraper: Recalls, Drug Labels & Adverse Events

Six FDA datasets behind one input form. Pull drug, device and food recalls, prescription and over-the-counter drug labels, FAERS adverse event reports, and the full National Drug Code directory — filtered by product, company, state, recall class and date, and flattened into spreadsheet-ready rows.

Mode	What you get
Drug recalls	FDA drug enforcement reports: class, reason, firm, distribution, dates
Device recalls	The same for medical devices
Food recalls	The same for food and cosmetics
Drug labels	SPL label text: boxed warning, indications, dosage, contraindications, interactions
Adverse events	FAERS reports: reactions, outcomes, suspect drugs, patient age and sex
NDC directory	Every listed drug product: NDC, labeler, dosage form, active ingredients

Why use this one

• The FDA's own API, nothing else. This reads api.fda.gov, the endpoint the Food and Drug Administration publishes for programmatic access. No HTML parsing, no login, no key required, and no terms-of-service grey area — so there is nothing to break when a web page is redesigned and no account to get blocked.
• Six datasets, one consistent output. Each dataset has its own field names and its own date field. This Actor translates one set of plain-English filters into the right query for whichever one you pick, so a date range means report_date for a recall and receivedate for an adverse event without you having to know that.
• Dates come out readable. openFDA stores dates as 20250131. Every date field here is ISO 2025-01-31, which sorts correctly in Excel and loads into a database as a date.
• The nesting is flattened. A FAERS report buries reactions and drugs under patient.reaction[] and patient.drug[]. You get reactions, reactionOutcomes, suspectDrugs and drugGenericNames as plain columns, plus counts. Label text sections come out as one column each.
• An escape hatch for power users. searchQuery passes a raw openFDA query clause through untouched and ANDs it with your other filters, so anything the form doesn't cover is still reachable.

How to use it

1. Choose What to collect.
2. Set any filters you want — leave them all empty to take the newest records in that dataset.
3. Click Start, then download results as JSON, CSV, Excel or HTML, or fetch them through the Apify API.

Example input: Class I drug recalls this year

{
  "mode": "drugRecalls",
  "classification": "Class I",
  "dateFrom": "2026-01-01",
  "maxResults": 1000
}

Example input: one company's recalls in California

{
  "mode": "foodRecalls",
  "firm": "Dole",
  "state": "CA",
  "maxResults": 500
}

Example input: serious adverse events mentioning a drug

{
  "mode": "drugEvents",
  "productSearch": "metformin",
  "seriousOnly": true,
  "dateFrom": "2025-01-01",
  "maxResults": 2000
}

Example input: every NDC product from one labeler

{
  "mode": "ndcDirectory",
  "firm": "Teva Pharmaceuticals USA, Inc.",
  "maxResults": 5000
}

Example input: a raw openFDA clause

{
  "mode": "drugLabels",
  "searchQuery": "openfda.route:\"TRANSDERMAL\" AND _exists_:boxed_warning",
  "maxResults": 200
}

Output

Each item carries a type field — recall, label, adverseEvent or drugProduct — and the dataset has a ready-made table view for each.

A recall

{
  "type": "recall",
  "dataset": "food",
  "recallNumber": "F-0123-2026",
  "eventId": "90123",
  "status": "Ongoing",
  "classification": "Class I",
  "productType": "Food",
  "recallingFirm": "Example Foods Inc",
  "firmCity": "Fresno",
  "firmState": "CA",
  "firmCountry": "United States",
  "productDescription": "Example brand spinach, 10 oz bag",
  "productQuantity": "12,480 cases",
  "reasonForRecall": "Potential Listeria monocytogenes contamination",
  "voluntaryMandated": "Voluntary: Firm initiated",
  "initialFirmNotification": "Press Release",
  "distributionPattern": "Nationwide",
  "recallInitiationDate": "2026-02-14",
  "reportDate": "2026-02-26",
  "terminationDate": null
}

An adverse event report

{
  "type": "adverseEvent",
  "safetyReportId": "12345678",
  "receiveDate": "2025-06-03",
  "serious": "1",
  "isSerious": true,
  "seriousnessHospitalization": "1",
  "reporterCountry": "US",
  "reporterQualification": "1",
  "reporterQualificationLabel": "physician",
  "patientSex": "2",
  "patientSexLabel": "female",
  "patientOnsetAge": 64.0,
  "patientOnsetAgeUnit": "801",
  "patientOnsetAgeUnitLabel": "year",
  "patientWeightKg": 78.0,
  "reactionCount": 2,
  "reactions": ["Nausea", "Dizziness"],
  "reactionOutcomes": ["1", "6"],
  "reactionOutcomeLabels": ["recovered/resolved", "unknown"],
  "drugCount": 3,
  "suspectDrugs": ["METFORMIN HYDROCHLORIDE"],
  "drugGenericNames": ["METFORMIN HYDROCHLORIDE"]
}

FAERS fields such as serious, patientSex, reporterQualification and reactionOutcomes keep the coded values the FDA publishes, exactly as received. Beside them, isSerious, patientSexLabel, reporterQualificationLabel, patientOnsetAgeUnitLabel and reactionOutcomeLabels translate them using the code lists in the openFDA field reference; a code outside those lists gets a null label rather than a guess. reactionOutcomes lines up one-to-one with reactions (null where the report gives no outcome), while drugIndications and drugRoutes list the distinct values across the report's drugs.

Pricing

Pay only for results:

Result	Price
Recall	$10.00 per 1,000
Drug label	$10.00 per 1,000
Adverse event report	$10.00 per 1,000
NDC drug product	$10.00 per 1,000

Set a maximum spend per run in the run options and the Actor stops cleanly when it's reached.

Limits to know about

These come from the FDA's API, not from this Actor:

• 26,000 hits per query. openFDA serves up to 1,000 records per page and refuses to page past 26,000 hits for one query. The Actor stops at that ceiling and tells you in the log. To collect more, split the work by date range — a year at a time, or a month for adverse events — and run it again.
• Rate limits. Without a key: 240 requests per minute and 1,000 requests per day, per IP address. With a free key from open.fda.gov: 240 requests per minute and 120,000 requests per day. Since each request brings back up to 1,000 records, the keyless daily allowance is already large; the optional openFDA API key input is there for heavy scheduled use.
• Adverse events are reports, not conclusions. A FAERS report says someone reported an event while taking a drug. It is not evidence the drug caused it, reports are not verified, and the same event can be reported more than once. The FDA is explicit about this, and so is every response the API sends.
• Adverse events have no company filter. FAERS reports carry a reporting company number but no searchable company name, so the Company field is rejected for that mode with a message rather than silently ignored. Search by drug name instead.
• openfda enrichment is sparse on recalls. The FDA attaches brand and manufacturer names to an enforcement report only when it can match the product to a listed drug. For food and device recalls that block is usually empty, so brandNames and genericNames are often [] — the product name you want is in productDescription.
• Label text is long. Each label section is truncated at 20,000 characters so a single label can't swamp your spreadsheet, and only the first block of each section is kept when the API returns several.
• Data lags. Each dataset has its own refresh schedule and the enforcement reports in particular are published weekly, so the newest recall in the API can be a few days behind an FDA press release.

FAQ

Do I need an account or API key? No. openFDA is open and keyless. A free key only raises the daily request limit, and it goes in the optional, secret openFDA API key field if you want it.

Can I monitor new recalls? Yes. Schedule the Actor daily with a dateFrom of the last week or so, and connect the dataset to Slack, Google Sheets, Zapier or Make.

Why are some fields missing from a row? openFDA omits a field entirely when a record has no value for it, so the row comes back with null rather than failing. Every field is read defensively.

What's the difference between drug labels and the NDC directory? The NDC directory is one row per marketed drug product — identifiers, dosage form, ingredients. A label is the full prescribing or Drug Facts text for an SPL document. Use the directory to enumerate products and labels to read what they say.

Is this legal? Yes. openFDA is a public US government API, published by the FDA for exactly this purpose, with no authentication. The FDA's own disclaimer accompanies every response: the data is not for clinical decision-making and should not be used to make decisions about medical care.

Feedback

Found a bug or need a field that's missing? Open an issue on the Actor's Issues tab.