Healthcare · pharmacy · lead generation · Medari Advisors · Delivered 30 Sep 2026

A lead finder for a healthcare accounting firm: one sentence in Claude, and back comes a checked sheet of real compounding pharmacies with the email each one publishes.

He grows his pharmacy email list state by state. Now he asks his own Claude for compounding pharmacies in a state and gets a sheet of real, independent ones with the email each publishes, and five random rows to spot-check.

1,046pharmacies that compound, on the starting lists for Oklahoma, Texas, Kansas and Arkansas, built from the national provider registry with chains, hospitals and specialty companies removed Counts on 2026-09-30, one row per phone number: Oklahoma 197, Texas 675, Kansas 99, Arkansas 75.

13 of 13 emails in the first real batch found on the pharmacy's own website, right business and town, by an independent check

Diagram: he asks his Claude for compounding pharmacy leads in Oklahoma; a starting list from the national registry, a Google Maps lookup and a match rule in code produce a sheet with one email per pharmacy, a call list and 5 rows to spot-check — One sentence in, a checked sheet out. The rules that decide are code.

What I built

The build

A starting list for each state, built by us from the national provider registry: every active pharmacy registered as compounding, with chains, grocery stores, national infusion companies, hospitals and clinics removed, one row per storefront, and the owner or pharmacist in charge as the contact.
A lookup in his own Apify account: one Google Maps search per pharmacy, reading the contact details its website publishes.
A match rule in code: a result counts only on a phone match, or a name and town match, so another business's email is never used.
A finished sheet: junk, web designer and head office addresses dropped, one email per pharmacy, his existing list deduplicated out, stores with no website kept as a call list with phone numbers, and 5 random rows to spot-check.

What changed

Before and after

Before

By hand

Finding pharmaciesSearched one at a time

A wrong matchEasy to miss

Checking the resultTrust the list

Cost per pharmacyHis time

After

The system

Finding pharmaciesOne sentence in his own Claude

A wrong matchDropped unless the phone, or name and town, match

Checking the result5 random rows to spot-check

Cost per pharmacyAbout half a cent

Where the numbers come from. The lookup was measured at $0.0054 a pharmacy, so Apify's free $5 a month covers about 900. In the first 13 real Oklahoma rows, 5 gave usable emails; rural stores often have no website and come back on the call list.

For engineers

How it's built

Architecture: A Claude plugin with one command and one skill. Everything that decides is code: the state lists are built ahead of time from the NPI registry, and the match and finish steps are Python scripts the skill runs. The model only drives the run and talks to him.
Where it runs: In his own Claude desktop app, with the lookup in his own Apify account. Nothing of ours in the loop.
Tests and evals: 32 script tests, 10 of them regression tests built from a real Oklahoma run. Anthropic's plugin validator passed on a clean unzip. A fresh agent ran it end to end following only the skill's instructions, and an independent check then confirmed every email it found.
Cost and latency: About half a cent a pharmacy in Apify, measured; the free $5 a month covers about 900.
Guardrails: A Google result counts only on a phone match, or a name and town match; otherwise its email is never used. The registry phone is primary and Google's is noted when it differs. Every batch ends with 5 random rows to spot-check.
The tradeoff we chose: Why this way and not the other
Apify instead of Clay: Apify Starter is $19 a month against $185 for Clay's cheapest paid plan (checked 29 Sep 2026). Apify's people add-on was tested and rejected: about 16 cents a pharmacy, and it matched staff from other businesses.

In their words

“This is great. Thanks for the explainer video, well done! I have installed the plug-in and it's working.”

Curtis Cole, CEO, Medari Advisors