MCP Server
Put NYC family data inside your assistant.
Motley speaks the Model Context Protocol — the open standard for connecting assistants to live data. Point a compatible client at our server and ask, in plain language, which NYC neighborhoods and schools fit your family.
What it is
The same dataset, as tools.
MCP is an open protocol that lets an assistant (Claude Desktop, and a growing set of other clients) call external tools mid- conversation. Our server exposes the data behind every Motley page — the 262 neighborhoods, ~5,800 schools, safety and health indicators, and family resources — as a small set of typed tools the assistant can query directly.
Every tool reads the same sourced, dated feeds documented on Our Data. No scraping our pages, no guessing — structured answers, straight from the source.
The tools
Eleven ways to ask.
- search-neighborhoodsSearch and filter NYC neighborhoods by borough or name. Returns NTA codes, names, boroughs, and centroids.
- get-neighborhoodFull profile for one neighborhood — demographics, raw scores, location, plus a family-readable interpretation (labeled score bands + cached blurbs).
- get-neighborhood-summaryDecision-ready summary of one neighborhood — a one-paragraph fit verdict, labeled score bands, and cached blurbs, with raw scores still available.
- get-schoolsSchools within a neighborhood — public, charter, and private — with quality ratings and grade levels.
- get-resourcesFamily resource counts — playgrounds, libraries, hospitals, recreation centers, farmers markets, ADA subway stations.
- get-activitiesKids' activities and businesses — classes, sports, arts, tutoring — from Foursquare data.
- get-safetySafety and health indicators — crime, collisions, housing violations, air quality, lead, asthma.
- get-comparableNeighborhoods most similar to a given one, by a 9-dimension score vector.
- get-admissions-methodsEvery NYC admission entry point — 3-K and Pre-K, Kindergarten, Gifted & Talented, and the 8 middle/high methods — each with its typical application window.
- explain-admissions-methodOne NYC admission entry point or method in full — the plain-English explainer plus its typical deadline window.
- get-admissions-demandHow competitive admission to one school is, program by program — seats vs. applicants, applicants-per-seat, and a competitiveness tier.
Connect it
We host it. Point your client at one URL.
No install, no database, no key — point any MCP-compatible client at our hosted endpoint:
https://mcp.motley.nyc/mcpMost clients take it from there. In ones with native remote-MCP support, open Settings → Connectors, add a custom connector, and paste that URL — the eleven tools appear, no restart required.
On a client that only loads servers over stdio through a config file, bridge to the same URL with mcp-remote instead. For Claude Desktop, add this to your claude_desktop_config.json and restart:
{
"mcpServers": {
"motley": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://mcp.motley.nyc/mcp"]
}
}
}Either way, ask something like “Compare Park Slope and Carroll Gardens for a family with young kids” — the assistant pulls the answer through the tools.
Advanced · run it yourself
Prefer to run the server on your own machine? It speaks stdio too. You'll need Node.js 22+ and your own DATABASE_URL for the data. Build it once:
cd mcp-server
npm install
npm run build # bundles to dist/index.jsThen point your client at the built file instead of the hosted URL:
{
"mcpServers": {
"motley": {
"command": "node",
"args": ["/path/to/motley/mcp-server/dist/index.js"],
"env": {
"DATABASE_URL": "<your-connection-string>"
}
}
}
}For providers
Run a tutoring or family service for NYC families? You can manage your Motley listing — draft it, verify, subscribe — from your own AI assistant too, over a private endpoint scoped to your account.
Open data, open protocol.
Prefer to explore it yourself? Everything the tools return is on the site — start with the neighborhoods, or see exactly where the numbers come from.