The directory is still for people
The DPC Finder has always been useful for humans.
You can look up direct primary care practices, search by location, browse states, and get a feel for who is practicing nearby.
But the way people use the internet is changing.
More and more, the first pass is not a person opening ten tabs.
It is an AI agent asking: "What should I look at?"
That changes what a directory needs to be.
Agents need structure, not scraping
A normal directory is human-readable. It has pages, filters, links, and listings.
An agent-native directory is different. It is structured. It is callable. It can answer targeted questions without scraping a bunch of pages and hoping the data comes out clean.
That is what we have started doing with the Tabflows MCP server.
The DPC Finder is now available through MCP, which means compatible AI tools can search the directory directly.
What agents can search
An AI agent can ask for:
- DPC practices in a state
- clinics in a specific city
- practices matching a name, doctor, specialty, or location
- listings by practice model
- full practice details by slug
That sounds technical, but the practical use is simple:
- A physician thinking about opening a DPC can ask:
Find DPC clinics within my state and summarize how crowded the market looks. - A consultant can ask:
Show me pure DPC practices in Tennessee and group them by city. - A clinic operator can ask:
Find practices near me that look similar to mine. - A researcher can ask:
Pull DPC practices in Texas and compare how many are listed in major metro areas.
The first pass gets faster
The old way was manual browsing.
The new way is: let the agent do the first sweep, then let the human make the judgment call.
That distinction matters.
We are not trying to make AI the expert on local healthcare markets.
We are giving AI a better map.
Better map, better questions
Most AI tools have weak context around DPC. They know the definition. They can explain the model. They can say "membership-based primary care."
But they do not automatically know the actual landscape: who is nearby, what the market looks like, which clinics exist, and how the ecosystem is spread out.
The DPC Finder already holds that kind of map.
The MCP server makes it usable by agents.
This is good for physicians. Good for clinic builders. Good for consultants. Good for anyone trying to understand the private-practice market without starting from a blank Google search.
The DPC Finder is still a directory for people.
Now it is also context your AI can use.