℞ TRY IT — OPENFDA MCP SERVER

Try it in 60 seconds

Want to see what an MCP server actually does? You can give Claude direct access to FDA drug data in about a minute. There's no coding and nothing to install.

What you'll need

A Claude account on any plan. Free accounts can add one custom connector.
About a minute on claude.ai or the Claude desktop app. Once it's set up, it works on your phone too.
On a company Team or Enterprise account? Your organization's owner has to add the connector.

Step 1 — Add the connector · about 30 seconds

In Claude's settings, find Connectors and choose to add a custom connector. Give it a name like OpenFDA and paste in this address:

https://openfda.datacompounded.com/mcp

Save it, and you're done with setup. Claude will note that custom connectors aren't verified by Anthropic. This one only reads public FDA data and never asks you to sign in.
Step 2 — Ask a question · about 30 seconds

Start a new chat and ask something in plain English. Try one of these:
- 💊“What does the FDA label say about using metformin in patients with kidney problems?”
- ⚠️“What are the most commonly reported adverse events for atorvastatin?”
- 🔍“Does the lisinopril label have a boxed warning? What does it say?”
- 🤰“What does the sertraline label say about use during pregnancy?”
Not seeing FDA data? Click the + in the chat box, open Connectors, and make sure OpenFDA is switched on.

What happens behind the scenes

Claude recognizes that your question needs real FDA data. It calls this server, which retrieves the official drug label or adverse event reports from the FDA's public database. Claude then reads the results and answers you. You may see a note that it used the OpenFDA tool. That's the MCP server at work.

That's the whole idea behind MCP: it gives an AI a safe, well-defined way to reach real data sources, instead of relying on what it memorized in training.

Why this matters

The same approach can connect an AI assistant to a pharmacy's formulary, a drug interaction database, inventory systems, or internal clinical references. The data stays authoritative and up to date, and the AI does the explaining.

Curious how the server was built, and the clinical judgment calls behind it? Read the case study.

Note: Results come from the FDA's public openFDA database and are provided for informational and demonstration purposes only. They aren't a substitute for professional clinical judgment. Adverse event reports reflect what was reported and don't establish that a drug caused an event.