Md Dmv MCP Server

@pipeworx/md-dmv

Maryland DMV MCP — monthly registered-vehicle counts and electric/plug-in hybrid adoption by county and ZIP code, from the Maryland Motor Vehicle Administration (MVA).

Tools

• md_dmv_vehicle_registrations(county?, month?, limit?) — registered vehicles by county for one month, with a genuine statewide total and each county's share of it. Answers "how many vehicles are registered in Montgomery County Maryland", "which Maryland county has the most registered vehicles", "how many cars are registered in Maryland".
• md_dmv_ev_adoption(county?, zip?, month?, group_by?, limit?) — battery-electric and plug-in hybrid registrations by county or by ZIP code for one month, with a Maryland total for context. Answers "how many EVs are registered in Montgomery County Maryland", "EV registrations in ZIP 20852", "which Maryland ZIP codes have the most EVs".

Auth

Keyless.

Data sources

• https://opendata.maryland.gov/resource/db8v-9ewn.json — MDOT MVA registered vehicles by county, monthly. Fields: year_month ("2026/06"), county, vehicle_count. 25 rows per month (23 counties, Baltimore City, and a NULL bucket for registrations the MVA could not assign).
• https://opendata.maryland.gov/resource/qtcv-n3tc.json — MDOT MVA electric and plug-in hybrid registrations by county, monthly. Fields: year_month, fuel_category ("Electric" | "Plug-In Hybrid"), county, count. ~464 rows per month.
• https://opendata.maryland.gov/resource/tugr-unu9.json — same, by ZIP code. Fields: year_month, fuel_category, zip_code, count. ~2,065 rows per month.

All three run monthly from 2020/07 to 2026/06 (latest as of 2026-07-29).

Gotcha: the Maryland WAF rejects query shapes, not queries

opendata.maryland.gov sits behind a Cloudflare WAF that answers certain SoQL query shapes with a "Just a moment..." interstitial — HTTP 403 with an HTML body — instead of data. It is reproducible, and it is specific to this one domain: data.ny.gov, data.wa.gov and data.ct.gov all accept the identical clauses without complaint. It was first hit from a deployed Cloudflare Worker, and has since been reproduced from a laptop too, so it is a property of the request, not of the caller's IP.

Confirmed to trip it:

Confirmed to work:

So this pack sends exactly one condition, pulls the whole month (at most ~2,100 rows), and does all county/ZIP narrowing, aggregation and sorting in code. That is cheap, and it is the difference between the pack returning data and returning an HTML challenge page. Do not "optimise" the filtering back onto the server.

A side benefit: because the complete month is always in hand, statewide_total_vehicles and maryland_plug_in_vehicles are real totals rather than sums of whatever survived limit. The row-level sum is reported separately as sum_of_returned_rows.

Gotcha: the same county is spelled two different ways

The registration file and the EV files disagree on punctuation for the same places:

Both tools fold punctuation away before matching, so a caller can pass either spelling.

Gotcha: the EV files carry out-of-state counties

Maryland records the registrant's county (or ZIP) of residence, so qtcv-n3tc carries ~300 non-Maryland counties with small counts — ADAMS, ALAMEDA, ALBEMARLE and so on — mixed in with the 24 Maryland jurisdictions. Every county row is flagged in_maryland, and maryland_plug_in_vehicles counts only the Maryland ones.

The nastiest instance: Virginia's PRINCE GEORGE sits right next to Maryland's PRINCE GEORGES, and Virginia's is the exact match for the phrase "Prince George". The county resolver therefore returns both rows (flagged) rather than silently picking the one-vehicle Virginia county over the 17,899-vehicle Maryland one.

Verified figures (2026/06, checked 2026-07-29)

Quick Start

Add to your MCP client (Claude Desktop, Cursor, Windsurf, etc.):

{
  "mcpServers": {
    "md-dmv": {
      "url": "https://gateway.pipeworx.io/md-dmv/mcp"
    }
  }
}

Or connect to the full Pipeworx gateway for access to all 1392+ data sources:

{
  "mcpServers": {
    "pipeworx": {
      "url": "https://gateway.pipeworx.io/mcp"
    }
  }
}

Using with ask_pipeworx

Instead of calling tools directly, you can ask questions in plain English:

ask_pipeworx({ question: "your question about Maryland MVA data" })

The gateway picks the right tool and fills the arguments automatically.

More

• All tools and guides
• pipeworx.io

License

MIT