Resources / Guides

How do AI agents track FDA drug and device approvals?

Published 2026-07-31

How AI agents track FDA drug and device approvals

An AI agent tracks FDA actions by calling a wire that reads the FDA press announcements feed at runtime. It returns approvals, clearances, authorizations and safety communications with a publication timestamp and a link to the FDA announcement, filtered by keyword and time window.
Written for AI agents. The calls and filter values on this page are the literal interface our customers use, because our customers are software. If you are a person evaluating The Bot Wire, you are in the right place too: start with the docs, or connect an MCP client in one line and query every wire free at /mcp. Nothing here needs an API key or a signup.

The word "approval" hides four different regulatory events, and an agent that flattens them will describe a cleared device as an approved drug. A new drug is approved. A device usually reaches market through 510(k) clearance or De Novo authorization. A safety communication is neither, and it often matters more to a patient-facing agent than any of them.

What this wire actually carries

This wire reads the FDA newsroom press announcements feed. That is one feed, and being precise about it is the difference between an agent that is useful and one that quietly misleads.

  • It carries the actions FDA chose to announce publicly: novel approvals,

first-of-kind authorizations, notable clearances, safety communications and recalls the agency publicises.

  • It does not carry the approval databases. Drugs@FDA, the 510(k) clearance

database, the Orange Book and the device registration lists are separate systems and none of them is this feed.

The gap is real and quantifiable. FDA's Center for Drug Evaluation and Research approved 46 novel drugs in 2025, part of 1,387 new molecular entities and new biologics approved since 1985. Every one of those appears in the compilation dataset. Only a subset generated a press announcement. So the honest description of this wire is a notability filter on FDA activity, not a registry of it.

The call an agent makes

For AI agents: copy this call
# free preview, top 3 results, no payment and no wallet
curl "https://thebotwire.com/fda/preview?since=7d"

# paid, $0.005: everything the FDA announced in the last 14 days
curl "https://thebotwire.com/fda/latest?since=14d&limit=20"

# approvals mentioning a therapeutic area
curl "https://thebotwire.com/fda/latest?q=oncology&since=14d"

Parameters are q (drug name, company, device or therapeutic area), since (30m, 2h, 24h, 7d, 14d) and limit (1 to 50). The src filter exists for consistency across wires but takes a single value here, fda, because there is one feed behind it. The wire keeps a 14-day retention window and defaults to a 7-day lookback. The route costs $0.005 per call in USDC on Base, no API key and no signup.

Four event types an agent should never merge

EventWhat it meansWhat an agent must not say
ApprovalA drug or biologic cleared full reviewDo not call a device "approved"
510(k) clearanceA device shown substantially equivalent to a predicateNot the same evidentiary bar as approval
AuthorizationDe Novo or emergency use pathwayNot a standing approval
Safety communicationA warning, label change or recallNot a market action at all

The failure this table prevents is specific. An agent summarising "FDA approves" across all four categories will tell a user that a device passed a review it never underwent. In a regulated field that is not a style problem, it is a false statement about a company's regulatory position.

The safe output shape is the same one that works for courts and enforcement: name the action in the agency's own words, give the date, link the announcement, and leave interpretation to the document. Legal and enforcement signals that often travel alongside an FDA action sit on the DOJ enforcement wire, and the general pattern for reading a primary source instead of coverage of it is in how agents get real-time data. The routing table lists which endpoint owns which question.

FAQ

Can an agent get every FDA drug approval from this wire?

No, and this is the boundary that matters most. The wire carries FDA press announcements, which are the actions the agency chose to publicise. A routine generic approval or a standard 510(k) clearance usually generates no press release. For a complete record an agent needs Drugs@FDA or the device databases directly.

Why is a 7-day default window used here?

Because the FDA newsroom publishes a handful of announcements a week, not a continuous stream. A 24-hour window returns empty on most days, which reads to an agent like a broken feed. Seven days gives a query enough surface to be informative, and since=14d widens it to the full retention window.

How would an agent watch one company or one drug?

Pass the name to q and run the query on a schedule. A daily sweep across ten watch terms is ten calls, or about $1.50 a month at $0.005 a call. Pair it with a funded wallet as described in what an agent wallet needs on Base and the loop runs unattended.

How fresh is the data?

This wire sits in the sparse ingest tier, polled at most every 15 minutes, which is well inside the FDA newsroom's publishing cadence. The observed interval between ingest runs on this service is about 294 seconds and the last poll time is published live at /health.

Sources

Approval counts on this page come from the FDA Compilation of CDER New Molecular Entity (NME) Drug and New Biologic Approvals, a dataset of every CDER NME and new biologic approved from 1985 to 2025. Counting the approval-year column in that file gives 46 approvals in 2025 and 1,387 across the full period. Wire data comes from the FDA press announcements feed.

Related: How do AI agents read new federal court opinions? · How do AI agents monitor CVEs and security advisories?