Health Newsletter Turns Its Own Archive Into a Searchable AI Tool

October 1, 2026 - 12:44

For someone who spends so much time discussing APIs and data exchange, this publication has had surprisingly little interoperability of its own. It has been old school software as a service, where the interface is the only way in.

Some of that comes down to the platform. Substack, despite its popularity, remains frustrating as a system of record. It optimizes for the reader, whose attention feeds the app, the Notes feed, and the recommendation network the company actually sells. That is the same trap Medium fell into years ago, just at a slower pace. Medium once attracted publications like The Ringer and Signal v. Noise, then lost them as it rebuilt around a paywall and an algorithmic feed. Substack's original pitch in 2020 was the antidote, since writers own their lists, but every new feed feature chips away at that promise.

Writers are generic supply in the eyes of the product team, and generic supply follows demand. So every product cycle pushes toward discovery: Notes, recommendations, the app, live video. The result is lackluster author tools, including laggy and broken analytics, all at a premium price.

The archive is where this stops being an author gripe and starts affecting readers. Across Substack and LinkedIn, this publication has put out 1,218 posts and nearly 900,000 words since 2018. Substack's native search can only see the 464 posts hosted there, and even then it struggles to surface relevant content. Readers hit the same wall.

Fixing search turned out to be far from trivial. There is no public API for readers to access content. RSS only carries the 20 most recent posts and cuts paid ones off at the preview. The only Substack MCP is an admin tool for analytics. The author export exists, but it is slow and awkward to work with.

So the Health API Guy MCP was built to work around all of that. It is a private MCP server holding the full text of everything published across Substack and LinkedIn. Add it to Claude or ChatGPT as a connector and you can ask questions directly. It pairs keyword search, useful for acronyms and case names, with semantic search. Every post is also labeled by topic, form, and court case, so a query can pull an entire lawsuit's docket across both platforms at once.

There is a bit of eating your own cooking here. The year has been spent arguing that headless healthcare is arriving and that developers want primitives over abstractions. Watching Salesforce plug its platform into Claude reinforced the bet that the records underneath outlive whatever interface sits on top. Keeping a personal archive behind a Substack search bar felt hypocritical by comparison. The archive is the body. Substack is one head, and whatever tool a reader uses is another.

The MCP is available to Founding Member subscribers for a few reasons. It costs money and time to run, with hosting, the database, authentication usage, and AI compute for search and labeling adding up as usage grows. Substack offers no way to automate access, since there is no API to check whether someone is a paid subscriber, so every user is manual operations for now. It is also a beta, and parts will break. Better that they break for a small group of dedicated readers who will report back. The Founding tier also deserved a real benefit. Until now it was mostly a generous tip.

Founding membership is now a flat $200 a year, replacing the old suggested $150 that readers could adjust. Existing Founding Members keep access grandfathered in at their previous rate. Larger teams can get access too, since group subscriptions at the Founding tier come with the group discount.

September's output covered a lot of ground. Articles included a case for agent native rails with intelligence at the endpoints, and a breakdown of interoperability as a series of workflows in a trenchcoat, complete with a color coded map showing which provider workflows went digital, which are fragmented, and which still run on fax.

Video content featured a ninety minute webinar on the FTC's investigation into Epic and why antitrust struggles with systems of record, plus a roundtable debate on whether Oracle Health is dead, who would actually buy Cerner, and why the front desk still hands you a clipboard.

On the regulatory side, a piece examined the CDC's request for information on public health data intermediaries and where AI could help a network handle the next emergency without another interface project.

Court coverage was heavy. Particle v. Epic advanced past an early summary judgment bid. The Epic discovery files revealed generative AI used for document review and denied interoperability requests. Epic v. Health Gorilla put fresh discovery on the public docket. Veeva v. Epic went to the Wisconsin Supreme Court. Texas v. Epic exposed unredacted filings and a $5 million document review budget. OpenEvidence v. Doximity raised the prospect of Rule 37 sanctions. And a roundup looked at Doximity's move to end OpenEvidence's case alongside a long sealed