The FDA inspection data API.
Form 483s, EIRs and warning letters as clean JSON: every observation's citations and investigator, plus FDA inspections, recalls, import alerts and registrations. Summaries and observation previews for every document.
- FDA document types
- 7
- search filters
- 30+
- 21 CFR section cited, linked
- Every
- inspection records
- Since 1996
Every Form 483, EIR and warning letter, as structured data
We read each page so your code doesn't have to.
From scanned page to fields
Firm, FEI, dates, investigators and each observation with its CFR citation, extracted and linked.
- firm_nameExample Pharma, Inc.
- fei3001234567
- inspection_end2024-03-14
- investigator_namesJ. Example
- observations[0].cfr211.192
- observations[0].topicsproduction_records
- classificationVAI
Seven document types
One schema across all of them.
Profiles built in
Firms, FDA investigators and regulations, each linked to its documents.
Filter on anything
FEI lookup, investigator, CFR section, classification, topic, place and date.
Every document, summarised
The summary, every extracted field and the opening words of each observation, with the document's link.
Simple REST, predictable JSON
Bearer API keys, read-only GET requests and paged results with a next_offset, so nothing is cut off.
- GET
/documentsSearch with every filter - GET
/documents/{id}All fields, observations, citations - GET
/call/facility_signalA site's FDA standing by FEI - GET
/call/search_recallsRecalls, import alerts, 510(k), NDC … - GET
/firms/{id or FEI}Site profile and history - GET
/investigators/{id}Investigator profile - GET
/observations483 observations and violations - GET
/cfr/{citation}A 21 CFR section and who was cited
What openFDA doesn't include
openFDA covers adverse events, recalls and labeling. Inspection documents are not part of it.
| Data | FDA Intel API | openFDA |
|---|---|---|
| Form 483 observations | Each one, with its CFR citations | Not included |
| EIR summaries | With the findings in brief | Not included |
| Warning letter violations | With close-out status | Not included |
| FDA investigators | Named, with profiles | Not included |
| Recalls and registrations | Yes, linked to sites | Yes |
| Adverse events, labeling | Not included | Yes |
Many teams use both: openFDA for product safety data, FDA Intel for what happened during inspections.
Need full document text at scale (thousands of documents) through an API? Talk to us.
Plans
The REST API, Claude and ChatGPT are included in both plans. Yearly billing saves 12%.
Standard
1 user$78/month
On the annual plan, billed yearly as $935Save 12%
FDA inspection data and document search for one person.
- Inspections and outcomes, 21 CFR citations, warning letters, import alerts, recalls, registrations and approvals
- Firm, site and investigator profiles, document search and observation previews
- In Claude, ChatGPT and through the API
10,000 requests a month (up to 1,000 a day).
Start Standard, yearlyPremium
Up to 3 users$218/month
On the annual plan, billed yearly as $2,615Save 12%
The same data for a small team, with five times the requests.
- Everything in Standard
- Up to 3 people on one subscription
- Five times the requests
50,000 requests a month (up to 5,000 a day).
Start Premium, yearlyEnterprise, larger teams or bulk data: talk to us.
One request is one tool call in Claude or ChatGPT, or one API call. Usage resets every month. On yearly billing the requests are still a monthly allowance: it resets every month of the year, on the same day of the month as your start date. A top-up adds 10,000 requests for $29, one time. Unused top-up requests end with the current monthly allowance period.
The plans cover data, summaries and observation previews. Full documents are bought and downloaded separately on the site.
Need full document text at scale (thousands of documents) through an API? Talk to us.
Cancel any time in Account → AI & API: Cancel plan keeps access until the end of the period you paid for, or use Manage billing (the Creem billing portal). Removing the connector from Claude or ChatGPT does not cancel the subscription.
Questions, answered
What is the FDA Intel API?
A REST API over the FDA Intel catalog of FDA inspection documents: Form 483s, establishment inspection reports (EIRs), warning letters, untitled letters, close-out letters and responses. It returns the structured fields extracted from each document as JSON over HTTPS, plus FDA public data: inspections and citations, warning-letter status, import alerts, recalls, drug shortages, registrations and approvals. Document records carry a summary, the opening words of each observation and the document's link; full documents are bought on the site.
Is there an FDA 483 API?
FDA Intel offers one. You can search Form 483s by firm, FEI number, investigator, date, place or cited regulation, and read the opening words of each observation with its 21 CFR citations. Plans start at $78 a month on the annual plan ($89 billed monthly).
Is there an FDA warning letter API or import alert API?
Yes, in the same API: warning letters with their violations, CFR citations and close-out status, FDA's warning-letter index, and import alerts with the firms on each red, yellow and green list, searchable by firm or FEI.
Does openFDA include Form 483s or EIRs?
openFDA publishes structured datasets such as adverse events, recalls, enforcement reports and product labeling. It does not include the text of Form 483s or EIRs, or the individual inspection observations in them. The FDA Intel API is built around those inspection documents.
How much does API access cost?
API access is part of both FDA Intel data plans, together with the Claude and ChatGPT connector. Standard is $78/month on the annual plan (billed yearly as $935) or $89/month billed monthly, for 1 user and 10,000 requests a month (up to 1,000 a day); Premium is $218/month on the annual plan (billed yearly as $2,615) or $249/month billed monthly, for up to 3 users and 50,000 requests a month (up to 5,000 a day). Yearly billing saves about 12%. Both plans are flat prices for the same data and include Claude, ChatGPT and the API. On yearly billing the requests are still a monthly allowance: it resets every month of the year, on the same day of the month as your start date. A top-up adds 10,000 requests for $29, one time. Unused top-up requests end with the current monthly allowance period. Documents are bought separately. Enterprise, larger teams or bulk data: talk to us.
Are there rate limits?
Yes, and they are published. One request is one tool call in Claude or ChatGPT, or one API call. Usage resets every month. Standard includes 10,000 requests a month (up to 1,000 a day); Premium 50,000 requests a month (up to 5,000 a day). Every reply carries X-RateLimit-Remaining and monthly usage headers, and your account page has a live meter. A top-up adds 10,000 requests for $29, one time. Unused top-up requests end with the current monthly allowance period. Long result lists are paginated with an offset and a next_offset value, so nothing is cut off silently.
Does the API include the full text of documents?
No. You get every structured field, the summary and the opening words of each observation, with the document's link. Full documents are bought and downloaded on the site. Need full document text at scale (thousands of documents) through an API? Talk to us.
How fresh is the data?
New documents are added to the catalog as they are processed. Every record carries its issue and inspection dates, and the stats endpoint reports catalog totals so you can track what changed.
What does the catalog cover?
Tens of thousands of FDA inspection documents for drug, biologics, device, food and other FDA-regulated sites in the United States and abroad, linked to the firms, FEI numbers, investigators and regulations they mention.
Can I look up a site by FEI number?
Yes. Pass an FEI to the documents or firms endpoint to get every document for that site, the investigators who inspected it, its inspection classifications and the CFR sections cited most.
Can I get FDA investigator data?
Yes. Investigator profiles list the documents and sites each investigator is named on, their co-investigators, how many observations they write on average, and the regulations and topics they cite most.
Is there a connector for Claude or ChatGPT?
Yes. The API is part of the AI connector subscription: the same data as an MCP server, so you can ask about inspection history in plain English inside Claude, ChatGPT or any MCP client. See the AI connector page.
Where is the API documentation?
The full reference is at /api/docs: authentication, rate limits, errors, every endpoint with parameters and examples in curl, JavaScript and Python, and an OpenAPI 3.1 spec at /api/openapi.json.