When to use which portal

DRAFT. First pass, 2026-07-06. Coverage descriptions below are accurate as of this writing but not exhaustively verified against live catalogs — treat as a starting map, not a final index.

New York State and New York City each run their own general-purpose open data catalog, and NYC additionally has several purpose-built government data sources that a general catalog search won’t surface well. This guide is a decision map: given a question, which source actually has the answer.

The two general catalogs

Portal Domain Platform What it covers
NYC Open Data data.cityofnewyork.us Socrata The City of New York’s full dataset catalog — thousands of datasets spanning parks, business licenses, demographics, permits, transportation, health, education, and more. The broadest net for “does the city publish data on X.”
NYS Open Data data.ny.gov Socrata New York State’s dataset catalog — state agency data: transportation, health, education, environment, economic development, and state government operations.

Both are reachable through the same socrata-mcp-server tools (search, get_data, fetch) — just point domain at the portal you need. See mcp-configs/socrata-nyc-nys.mcp.json for pre-wired aliases, and query-patterns.md for the actual query syntax.

Use a general catalog when: you don’t know if a specific purpose-built tool exists for your question, you need to browse/discover datasets by keyword or category, or the data you need doesn’t fit any of the specialized sources below.

NYC’s purpose-built sources (BetaNYC MCPs)

These cover NYC domains that a keyword search in the general catalog handles poorly — legislative process, procurement, government spending, service requests, and legal text are structured very differently from a typical open dataset.

Source MCP What it covers Why not just search the catalog
NYC Council (Legistar) nyc-council-mcp Bills, hearings, votes, committees, council members Legislative data is relational (a bill has sponsors, committee history, vote records) — the catalog has flat tables, not this structure.
NYC City Record nyc-record-mcp Procurement notices — solicitations, contract awards, public hearings Procurement notices are published as a daily record, not a queryable dataset.
NYC Checkbook nyc-checkbook-mcp Spending, contracts, budget, payroll, revenue (via NYC Comptroller’s API) The Comptroller’s Checkbook API has its own schema and update cadence, separate from the Open Data catalog.
NYC 311 nyc-311-mcp Service-request lookup, city-services calendar, emergency/weather alerts 311’s live API answers “what’s the status of this specific request” — the catalog’s 311 dataset is a bulk historical export, not a live lookup.
NYC Charter, Administrative Code, Rules nyc-charter-laws-rules Full legal text (offline corpus) This is legal text, not tabular data — full-text search across chapters and sections, not a dataset schema.

NYS’s purpose-built source

Source MCP What it covers Why not just search the catalog
NYS Open Legislation nys-openlegislation-mcp Bills, laws, members, committees, calendars, agendas, floor/hearing transcripts (150k+ records) Same shape as NYC Council — legislative process data is relational, not a flat dataset.

Decision guide

  1. Is the question about a specific legal/legislative/procurement/spending/service-request process? → Use the matching purpose-built MCP above.
  2. Is the question “does the city/state publish data on X” or “find me a dataset about Y”? → Search the general Socrata catalog (NYC or NYS, per the question’s jurisdiction).
  3. Not sure which government level has it? → Search both catalogs; NYC and NYS often publish overlapping-but-different data (e.g., both publish some environmental and health data, at different granularity).
  4. Looking for county or municipal data outside NYC? → See resources/ny-open-data-portals.md — most NY counties and cities run their own separate portal, not part of either catalog above.

A note on trust

Every source table above traces back to an authoritative government system — Legistar, the Comptroller’s Checkbook, the 311 system, the Charter itself, or the Socrata-hosted catalogs. That traceability is the point of this whole repo: an AI agent citing “NYC Checkbook, contract #12345” is verifiable; an agent guessing from training data is not. See grounding-ai-agents.md for why this matters and how to build for it.