openFDA Scraper
Pricing
from $0.97 / 1,000 records
openFDA Scraper
openFDA scraper: one row per FDA record, from drug, device and food recalls, marketed drug products in the NDC directory, or 510(k) device clearances. Pick one dataset per run and narrow it with FDA's search syntax and dates. No account. $1.00 per 1,000 records.
Pricing
from $0.97 / 1,000 records
Rating
0.0
(0)
Developer
Dami's Studio
Maintained by CommunityActor stats
0
Bookmarked
2
Total users
1
Monthly active users
6 days ago
Last modified
Categories
Features
Share
openFDA scraper for one FDA dataset at a time, one flat row per record: drug, device or food recalls, the NDC drug directory, or 510(k) device clearances, narrowed with FDA's own search syntax and a date range. Each row carries FDA's own id, the fields that dataset holds, and a link that returns the untouched original.
The part to know first: FDA itself says these records should not drive public alerts.
| Input | One dataset, an optional FDA search expression and a date range |
| Output | One row per FDA record: a recall with its firm, product, hazard class, reason and dates; a marketed drug product; or a 510(k) clearance with its device, applicant and decision |
| Ceiling | 30,000 records per run, one dataset per run |
| Account needed | None. No FDA API key |
| Price | $1.00 per 1,000 records on the Free plan, a little less on paid plans. The free plan's $5 a month covers about 5,000 |
🔍 What openFDA Scraper does
Each run reads one dataset and returns one row shape. A recall, a marketed product and a premarket
clearance share almost no fields, so they never land in the same table. The three recall datasets do
share one field list, and three runs stack into a single spreadsheet, with the dataset column
saying which run each row came from.
dataset | One row is | The date range applies to | FDA refreshes it |
|---|---|---|---|
drug-recalls | one drug enforcement action | report date | weekly |
device-recalls | one device enforcement action | report date | weekly |
food-recalls | one food enforcement action | report date | weekly |
drug-products | one marketed product in the NDC directory | marketing start date | daily |
device-clearances | one 510(k) premarket clearance | decision date | monthly |
Records come newest first. Repeated records are dropped before you are billed.
When a search matches nothing, a free row tells you whether nothing matched or a field name in your query does not exist.
📋 What data you get from each FDA record
| What you get | Field |
|---|---|
| On every row: the dataset, the kind of record, FDA's own id, and a link to the untouched original | dataset, recordType, recordId, apiUrl, scrapedAt |
| Recalls: the firm, the product and how much of it, the hazard class, the reason and the status | recallingFirm, productDescription, productQuantity, codeInfo, classification, reasonForRecall, status, productType, eventId |
| Recalls: how it was started and where the product went | voluntaryMandated, initialFirmNotification, distributionPattern |
| Recalls: the firm's location | city, state, country, postalCode |
| Recalls: the dates | recallInitiationDate, centerClassificationDate, reportDate, terminationDate |
| 510(k) clearances: the device, the applicant and its codes | kNumber, deviceName, applicant, deviceClass, productCode, regulationNumber, medicalSpecialty, advisoryCommittee |
| 510(k) clearances: the decision and its dates | clearanceType, decisionCode, decisionDescription, decisionDate, dateReceived, expeditedReview, thirdParty, statementOrSummary |
| Drug products: the NDC directory's fields, including the ingredients, the route and every package code | activeIngredients, route, packageNdcs, packageCount |
▶️ How to scrape FDA recalls and device clearances
- Open openFDA Scraper and click Try for free.
- Pick a dataset in Which FDA records. The form starts on drug recalls.
- Type a query into Search expression (optional), or clear it for the newest records. It arrives
set to
classification:"Class I". - Set From date (optional), To date (optional) and Maximum records.
- Click Start, then open the dataset view for your run and download it as CSV, JSON or Excel.
💰 How much does it cost to scrape openFDA?
$1.00 per 1,000 records, which is $0.001 a record on the Free plan. Paid plans pay a little less: $0.99 on Starter, $0.98 on Scale and $0.97 on Business. On the free plan, the $5 Apify gives you each month covers about 5,000 records. The 100-record example below comes to ten cents.
You pay for record rows and nothing else per row. The sample row and every diagnostic row are free, and so are the repeats FDA hands back twice. A run stopped on an unreadable date asks FDA for nothing. If you set a maximum charge on the run, it stops at that figure instead of going past it.
📥 What you give it
{"dataset": "drug-recalls","search": "classification:\"Class I\"","dateFrom": "2026-01-01","maxItems": 100}
| Field | Default | What it is |
|---|---|---|
dataset | none, which gives the free sample row; the form starts on drug-recalls | drug-recalls, device-recalls, food-recalls, drug-products or device-clearances. Empty, or any other value, returns the free sample row instead of records. |
search | none, which gives the most recent records; the box starts with a Class I search | FDA's query syntax, such as recalling_firm:"Abbott", classification:"Class I", brand_name:"Tylenol" or device_name:"catheter". Join clauses with +AND+ or +OR+, quote phrases, and add .exact to a field name to match the whole value. A bare "salmonella" searches the whole record. |
dateFrom | none; the box starts at 2026-01-01 | Year first: 2026-01-01, 20260101, 2026/1/5 or 2026.01.05. Anything else, a full timestamp included, stops the run with a free BAD_INPUT row before FDA is asked. |
dateTo | none | Same formats. A range that runs backwards stops the run the same way. |
maxItems | 100 | 1 to 30,000 records, newest first. |
proxyConfiguration | leave it alone | Only if you want the traffic to go through servers of your own. |
📤 What you get back
A real row from a run on 30 September 2026:
{"ok": true,"charged": true,"dataset": "drug-recalls","recordType": "enforcement","recordId": "D-0866-2026","recallingFirm": "VITRUVIAS THERAPEUTICS INC","productDescription": "Thyroid Tablets, USP, 1/2 Grain (30 mg), Each tablet contains: levothyroxine (T4) 19 mcg, liothyronine (T3) 4.5 mcg, 100 Tablets, Rx only, Distributed by: Vitruvias Therapeutics, Auburn, AL 36830, Product of USA, NDC 69680-166-00.","classification": "Class I","status": "Ongoing","reasonForRecall": "Superpotent Drug","productType": "Drugs","eventId": "99709","distributionPattern": "US Nationwide.","productQuantity": "1955 bottles","codeInfo": "Lot # 504950, Exp Date: 09/30/2026.","voluntaryMandated": "Voluntary: Firm initiated","initialFirmNotification": "Letter","city": "Auburn","state": "AL","country": "United States","postalCode": "36830-4318","recallInitiationDate": "2026-08-21","centerClassificationDate": "2026-09-22","reportDate": "2026-09-23","terminationDate": null,"apiUrl": "https://api.fda.gov/drug/enforcement.json?search=recall_number:%22D-0866-2026%22&limit=1","scrapedAt": "2026-09-30T02:31:36.343Z"}
| Field | How to read it |
|---|---|
dataset, recordType | The dataset you asked for, and the grain of the row: enforcement, drug-product or device-clearance. |
recordId | FDA's own id: a recall number, an NDC product code or a K number. Stable, so use it as the key across runs. |
apiUrl | The FDA address that returns this one record untouched, including the fields left out here. |
classification | FDA's hazard class. Class I is the serious end: a reasonable chance of serious harm or death. |
status, terminationDate | terminationDate is null while the recall counts as open. |
activeIngredients, route | Drug products only. Kits have no single strength or route, so both are null on kits. |
packageNdcs, packageCount | Drug products only. Every package code under the product, and how many there are. |
deviceClass | Device clearances only. FDA's risk class, 1, 2 or 3, blank on older records. |
🧾 Reading the output
Three kinds of row can share a dataset, and charged is the field that separates them.
| Row | How to spot it | Charged |
|---|---|---|
| A record | charged: true | yes |
| The sample | _sample: true, charged: false | no |
| A diagnostic | _diagnostic: true, charged: false, an errorCode | no |
The sample row also says recordType: "enforcement", so filter on charged == true, not on
recordType.
| Code | What it means |
|---|---|
NO_RESULTS | The query was valid and FDA holds nothing matching it. |
BAD_INPUT | A field name in your search does not exist on that dataset, and the row names it. Also a date that could not be read, a backwards range, or a request FDA rejected. |
SERVER_ERROR | FDA refused the request several times over, which nearly always means the search expression could not be parsed. Check the brackets and quotes. |
NETWORK | FDA could not be reached. |
TIME_BUDGET | The run got close to its time limit and stopped on its own. The records delivered before that are complete and charged; records it never reached are not. The run still ends as succeeded, with an amber status message, even when the time limit stopped it before any record came back. |
UNEXPECTED_ERROR | The run stopped early on an unexpected error. The records before this row are kept. |
In the Console, switch the dataset to the view that matches your run: Recalls and enforcement, Drug products or Device clearances. The default Overview is only for a quick look.
💡 What people use it for
- Watching for recalls that touch your own supply chain:
drug-recallson a schedule with arecalling_firmor product term, compared onrecordId. - Keeping a formulary or product catalogue current with
drug-productsfiltered bylabeler_name. - Tracking who cleared which medical device and when, with
device-clearancesfiltered byproduct_codeoradvisory_committee. - Food-safety monitoring for a distributor or a grocery buyer, with
food-recallsnarrowed to a state.
🚧 What it does not do
- No adverse-event reports. Those describe individual patients, so they are left out on purpose.
- No drug labelling text. The prescribing sections are a different shape and far larger per record.
- No medical advice, and no alert service. FDA says these records should not be relied on for medical decisions or public alerting.
- No product matching. A recall carries FDA's free-text
productDescription, so tying it to your own SKU is still your job. - US FDA only. No other regulator is included.
- It is not an official openFDA tool. Dami's Studio is independent and is not affiliated with or endorsed by openFDA or the FDA, or by any other company named on this page.
🧭 Which public records scraper do you need?
| If you want | Use |
|---|---|
| FDA recalls, NDC drug products and 510(k) clearances | This one |
| Safety recalls and specs for a vehicle, by VIN | VIN Decoder & Recalls |
| Patents by keyword, inventor or assignee | Google Patents Search Scraper |
❓ Questions people ask
Do I need an FDA API key?
No. Nothing to sign up for, and the run never asks for a credential.
How many FDA records can one run return?
Up to 30,000. For more, split the date range across runs.
What happens if my search matches nothing?
One free row. NO_RESULTS means FDA genuinely holds nothing. BAD_INPUT means a field name in your
query does not exist on that dataset, and the row names it.
Can I combine drug, device and food recalls?
Yes. They share one field list, so three runs concatenate into one table.
Can I use this to send recall alerts to customers or patients?
No. FDA says so about the source itself. Use them for research and monitoring, and FDA's own announcements for anything that warns people.
Can I call it from code or connect it to an AI assistant?
The API tab has code for Python, JavaScript
and the CLI. MCP clients like Claude and ChatGPT use
https://mcp.apify.com/?tools=fetch-actor-details,dami_studio/openfda-drug-device-scraper; coding assistants can load
our skill
file. All three
bill your Apify account at the same price.
Is it legal to use openFDA data?
The records come from openFDA, the FDA's public API. FDA does not endorse this actor. Apify's write-up on the legality of web scraping covers the wider question, and we are not lawyers.
🆘 If something breaks
Open the Issues tab on the actor page and send the run ID with the input you used. The
errorCode and details on a diagnostic row usually name the reason already.