What is MCP? The short version

MCP stands for Model Context Protocol. It is a standard way to plug a live data source into an AI assistant, so the assistant can look things up for you instead of guessing from whatever it memorized during training. Anthropic published it as an open standard in late 2024, and it is now supported by Claude, ChatGPT, Cursor, and most serious AI tools.

The problem it solves

An AI assistant on its own is a very well-read person with no phone: it knows a lot, but it cannot check anything current. Ask it about a specific parcel and the honest answer is "I have no idea, my training data is not a county database." MCP gives the assistant a phone. When you connect an MCP server, the assistant gains a set of concrete actions it can take: search these parcels, fetch this owner, pull this sale history. It decides when to use them, makes the call, and reads the live answer before it replies to you.

What that looks like in practice: you type "find vacant lots over two acres in Rutherford County with out-of-state owners" into Claude. Claude sees it has a parcel-search action available, fills in the filters, calls the server, gets rows back, and writes you a summary with the actual properties. No copy-paste, no export-import, no stale data.

How is that different from an API?

It is not a rival; it is a layer on top. An API is the counter where programs place orders, and it expects exact phrasing. MCP is the standard that lets an AI assistant read the menu and do the ordering itself. You bring the plain English; the protocol handles the exact phrasing. That is why MCP matters for people who do not code: it makes every connected database usable by typing sentences.

If the word API itself is new, start with What is an API? - five minutes there makes this page obvious.

What connecting actually involves

Three steps, once: paste the server's address into your AI tool's connector settings, sign in to your own account on the data provider's page, approve the connection. From then on the assistant can use that source in any conversation, and the first time it wants to, it asks your permission.

Is it safe?

Security works the way you would hope. You connect a server once, sign in to your own account, and the assistant acts as you, with your permissions and your usage limits. Disconnect it and the access is gone. The assistant never sees your password; the sign-in happens on the data provider's own page, the same way "Sign in with Google" works. Three things follow from that design:

  • Your plan is the ceiling. The assistant can only see what your account tier can see, and everything it pulls is metered against your allowance like any other use.
  • Revocation is real. Remove the connector, or revoke it from your account page, and the access dies with it.
  • No shared secrets. There is no password pasted into a chat window, ever. If a tool asks you to do that, it is not MCP.

Try it with parcel data

The walkthrough in Use JB PARCEL with AI tools takes about two minutes and does not require writing any code.

Connect Claude to 151.0 million parcels.
Paste one URL, sign in, and ask questions in plain English. Not open for signup yet: join the waiting list and your first month of Professional is on us at launch.
See the setup