US Product Recalls Search API

What this Actor returns

Search US food, drug and device recalls in the official enforcement reports of the US food and drug regulator, and get one clean, typed record per recall. The records are made for a spreadsheet or a monitoring job: one row each, the same 12 fields every time, and a value the source does not have is null, never a missing key.

• recall_number, event_id, product_type: which recall it is (the number starts with F for food, D for drug, Z for device)
• classification, status, voluntary_mandated: how serious it is, where it stands and who started it
• recalling_firm_title, firm_city, firm_state, firm_country: the firm, as the regulator publishes the name
• recall_initiation_date, report_date: when the firm started the recall and when it appeared in the weekly report

You can filter by category, report date, class, status, a word in the firm's name and the firm's state. The Actor reads the regulator's public data API and nothing else. No login, no scraping.

It does not return the reason for the recall, the product description, the distribution text, addresses or device terminology codes (GMDN), and it does not say what to do about a recall. Rows without a recall number are left out.

Who it helps

• Quality, compliance and supply chain teams who watch recalls that touch their suppliers, products or categories, and want them in a table or a feed.
• Analysts, journalists and developers who want tidy recall data for a dashboard or a model, through the Apify API, without writing their own client for the source.

Quick start

1. Open the Input tab and leave the defaults as they are. They search all three categories for recalls reported in January 2024, so a first run returns 10 records (the default Maximum records).
2. Press Start. The default run costs about $0.03: 10 records at $0.003 each, plus Apify's tiny start event (see What it costs). Try it on one record first by setting Maximum records to 1; it costs a fraction of a cent.
3. Open the Output tab, check the records, then set your own categories, dates and filters.

Input

Field	What it does	Default
Categories	food, drug, device; one or more. They are read in that order.	all three
Reported from / to	Only recalls whose report date is in this range, as YYYY-MM-DD. Empty for no bound.	2024-01-01 to 2024-01-31
Class	Class I, Class II or Class III. Empty for any.	empty
Status	Ongoing, Completed, Terminated or Pending. Empty for any.	empty
Firm name contains	Words in the recalling firm's name, like Acme Foods. Letters, digits, spaces and . , & ' - only. Empty for any.	empty
Firm state	A two-letter code like CA. Empty for any.	empty
Your own openFDA key	Optional and secret. See Limits and honest notes.	empty
Maximum records	How many records to return at most, from 1 to 50000.	10

Records come category by category, in recall number order within each category.

Output

One record per enforcement report row. The record below is what the default input returned on 13 October 2026, unchanged (the first of the 10 records).

{
  "recall_number": "F-0655-2024",
  "event_id": "93478",
  "classification": "Class II",
  "status": "Terminated",
  "product_type": "Food",
  "recalling_firm_title": "Publix Super Markets, Inc.",
  "recall_initiation_date": "2023-11-16",
  "report_date": "2024-01-03",
  "voluntary_mandated": "Voluntary: Firm initiated",
  "firm_city": "Lakeland",
  "firm_state": "FL",
  "firm_country": "United States"
}

What it costs

You pay for one event, Recall record, at $0.003 per record returned. Apify also adds its standard run-start event, $0.00005 per start for an Actor run with up to 1 GB of memory; it is the platform's own event, not ours. The default run of 10 records at $0.003 costs about $0.03005. Records the Actor leaves out (see below) are never charged.

For example, every Class I food recall in a year is a few dozen records, which costs a few cents. If your Apify plan includes monthly credit, a small run draws on that credit.

If the source cannot be reached or keeps answering with errors, the run stops with a clear status message, and only the records already returned are charged.

Limits and honest notes

• Requests per day. The source allows 1,000 requests a day per address without a key, and 120,000 a day per key. This Actor runs without a key unless you give it one, and Apify runs share addresses, so a busy day may use up the keyless allowance. If a run ends with a message that the source is busy, wait and run again, or give your own free key in the optional key field (the run's message points there too); it is secret, it is sent only to the source, and it raises the limit to 120,000 requests a day.
• Pace. The source allows 240 requests a minute. The Actor sends at most 120 a minute, always says who it is in the request, and reads up to 100 rows per request; when the source is busy it waits as asked, up to about a minute per request, and then stops with a clear message.
• The reach of one search. The source lets a search skip at most 25,000 rows. A search with more matches than that returns the first part of each category and the status message says how many matches it could not reach; narrow the dates or add a filter to get the rest.
• Freshness follows the source, which publishes the enforcement reports weekly. report_date is the date of the weekly report, not the date the Actor ran.
• The firm name is published data. It is output as the regulator publishes it. A small firm's name can be a person's name; the Actor does not change names and returns no other personal details.
• Not a decision tool. The source itself says its data are unvalidated and not to be relied on for medical care decisions. Check a recall with the regulator before you act on it.
• What the filters match. Firm name matches whole words in the name. Class and status match the source's own wording.

Questions

Why did my run return fewer records than I asked for? Either fewer recalls matched, or some rows were left out because they lacked a recall number in the usual shape or a core field. You were not charged for those. The final status message counts them and says why.

What does "beyond_the_skip_limit" mean in the status message? The search matched more rows in that category than the source lets any search reach. The number says how many were out of reach. Narrow the dates and run again.

What does the run say when something is wrong? It says what it found and what to try. A message about a field in the input names the field to fix and shows the right shape. A message about the key means the source refused it: check the key, or leave it empty. "Could not be reached" or "kept answering" means the source is busy; wait a few minutes and run again.

Why does a search for a firm name return firms with other words too? The name filter finds your words anywhere in the firm's name, so Acme also finds Acme Foods Inc.

Why are there several records for one event? One recall event can cover several products, and each gets its own recall number. event_id ties them together.

Can I run it on a schedule? Yes, with the Apify scheduler or the API. Each run is independent and charged per record returned.

Data source and AI note

Data source: the openFDA enforcement reports (food, drug and device) of the U.S. Food and Drug Administration, released under the Creative Commons CC0 1.0 Universal licence, a public domain dedication that allows commercial use without asking permission. Data provided by the U.S. Food and Drug Administration. report_date is the date of the weekly enforcement report; the records have no source_updated_at field.

This tool is not affiliated with, endorsed by or sponsored by the U.S. Food and Drug Administration.

This tool was built with the help of AI coding assistants.

Help

Questions, a record that looks wrong, or an idea for a field: please open an issue on the Issues tab. Thank you for the report; every one is read.