Skip to main content
Glama

QueensEstate ♔⛫

CI

An MCP server that lets an LLM answer Charlotte residents' everyday questions with official public data from the City of Charlotte Open Data Portal — https://data.charlottenc.gov.

The portal lives at data.charlottenc.gov. It is an ArcGIS Hub site; its ~380 datasets are ArcGIS REST layers run by the City of Charlotte and Mecklenburg County. No API key is required.

What's on the portal

A survey of the catalog (/api/feed/dcat-us/1.1.json) shows these main themes:

Theme

Examples

Public safety

CMPD incidents (870k+ since 2017, updated daily), homicides, traffic stops, fire incident reports, crashes

City services

311 service requests (3.4M), solid-waste collection routes, code enforcement cases, street closures

Planning & zoning

Zoning, rezoning petitions, historic districts, 2040 policy map, floodplains, parcels

Transportation

CATS bus routes and stops, LYNX light rail, park-and-ride, bike lanes, sidewalks, greenways

Places

Libraries, schools, parks, fire/police stations, pharmacies, grocery stores, EV chargers, day care

Government

Capital projects, budget, employee salaries, council and commissioner districts

Demographics

Census block groups/tracts, Quality of Life neighborhood profiles

Related MCP server: mcp-arcgis-durham

Tools

The tools are chosen for the questions residents actually ask. Every tool is read-only, and every location parameter accepts either an address ("600 E 4th St, Charlotte") or "latitude,longitude".

My address

Tool

Answers questions like

get_address_profile

"Who is my council member?" "What police division and fire station serve me?" "What's my zoning? Am I in a flood zone or historic district?"

get_trash_and_recycling_schedule

"What day is trash pickup? Is this a recycling week (GREEN/ORANGE)?"

lookup_address

"Is this a valid address? What's its parcel ID and municipality?"

Safety

Tool

Answers questions like

get_crime_near

"Any car break-ins near my apartment this month?"

summarize_crime

"How have robberies in Charlotte changed since 2019?" "Which division has the most burglaries?"

get_traffic_crashes_near

"Is the intersection by my kid's school dangerous?"

City services and neighborhood change

Tool

Answers questions like

get_311_requests_near

"Has anyone already reported this pothole or streetlight?"

get_code_enforcement_cases

"Does the house I'm renting have open housing-code violations?"

get_street_closures

"Are any roads closed on my commute?"

get_capital_projects_near

"What's being built near me, when is it done, and who do I contact?"

get_pending_rezonings

"Is anyone trying to rezone land near my neighborhood?"

Places and transit

Tool

Answers questions like

find_nearby_places

Nearest library, school, park, greenway, pharmacy, grocery store, EV charger, light rail station, bus stop, public Wi-Fi… (18 categories)

get_bus_route

"How often does the 9 run on Sunday evenings?"

Government transparency

Tool

Answers questions like

get_city_budget

"How much does the city budget for Police?"

get_city_salary_stats

"What do Charlotte firefighters earn on average?" (aggregates only, no names)

Anything else in the catalog

Tool

Purpose

search_datasets

Keyword search over the whole portal

describe_dataset

Fields, types, and record count for a layer

query_dataset

SQL-filtered records from any layer, optionally near a location

summarize_dataset

Server-side counts/sums grouped by fields

Setup

Requires Python 3.14 (developed on 3.14.7).

~/.pyenv/versions/3.14.7/bin/python -m venv .venv
.venv/bin/python -m pip install -e . --group dev

Use with Claude Code

claude mcp add queensestate -- /absolute/path/to/queensestate/.venv/bin/queensestate

Use with Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "queensestate": {
      "command": "/absolute/path/to/queensestate/.venv/bin/queensestate"
    }
  }
}

Other transports and debugging

.venv/bin/queensestate --transport http --port 8000        # http://127.0.0.1:8000/mcp
npx @modelcontextprotocol/inspector .venv/bin/queensestate  # interactive inspector

The http transport authenticates its callers with Google OAuth and refuses to start without a public URL, a Google client, and an allow list — see Hosting. To try it locally:

QUEENSESTATE_PUBLIC_URL=http://localhost:8000 \
QUEENSESTATE_GOOGLE_CLIENT_ID=... QUEENSESTATE_GOOGLE_CLIENT_SECRET=... \
QUEENSESTATE_ALLOWED_EMAILS=you@gmail.com \
  .venv/bin/queensestate --transport http --port 8000

Hosting

The hosted server runs on Google Cloud Run at https://queensestate.adamwanninger.com/mcp, and is its own OAuth 2.1 authorization server: it delegates the sign-in to Google purely to learn which account is calling, then checks that address against an allow list and mints its own tokens. Google's tokens are never passed through to the MCP client.

claude mcp add --transport http queensestate https://queensestate.adamwanninger.com/mcp

scripts/deploy-gcp.sh provisions and deploys the whole thing, and re-running it is also how you update. See deploy/GCP.md.

Hosting is only for the HTTP transport. Running the server locally over stdio needs no Google account, no allow list, and none of the cloud dependencies.

Development

.venv/bin/pytest                 # unit + in-memory MCP tests; no network access
.venv/bin/ruff format --check . && .venv/bin/ruff check .
.venv/bin/mypy

Layout:

  • src/queensestate/arcgis.py: async ArcGIS REST / Hub client (timeouts, bounded retries, host allowlist)

  • src/queensestate/address.py: address matching against the county Master Address layer

  • src/queensestate/catalog.py: the curated layers behind the purpose-built tools

  • src/queensestate/server.py: the MCP tools

Data notes and limitations

  • Geocoding uses Mecklenburg County's Master Address layer, so only addresses inside the county resolve. Coordinates outside the county are rejected.

  • Freshness varies by dataset: CMPD incidents, 311 requests, and code enforcement are current within days; the Budget Report dataset currently runs FY2018–FY2023. Tools report what the portal publishes.

  • CMPD locations are generalized by the department for privacy.

  • 311 data records when requests were received, not when they were resolved.

  • Timestamps are ISO 8601 in Charlotte local time with the UTC offset (e.g. 2026-09-06T00:00:00-04:00). The underlying services store UTC.

  • Budget amounts include large negative lines (e.g. "00 Non Department", Charlotte Water), so get_city_budget reports both the net total and the sum of positive lines.

  • Statistics limits: some city map services ignore resultRecordCount on aggregate queries, so the client enforces row limits itself.

  • Live bus arrivals are not on the portal; pair this server with a CATS real-time source. Bus stop IDs from find_nearby_places match CATS stop IDs.

  • The generic tools only reach hosts that serve portal datasets (gis.charlottenc.gov, meckgis.mecklenburgcountync.gov, services.arcgis.com, gis.ci.charlotte.nc.us).

  • Mecklenburg County's server rejects default Python user agents, so the client sends its own.

Ideas for more tools

  • Neighborhood Quality of Life indicators by NPA (income, age, housing, amenities)

  • Charlotte Fire incident reports by address block

  • CMPD officer traffic-stop statistics

  • Sidewalk and bike network gaps near an address

  • Tree canopy and land surface temperature by neighborhood

License

MIT - see LICENSE.

Available Tools

19 tools
describe_datasetDescribe a dataset's fieldsA
Read-onlyIdempotent
Inspect

Show a dataset layer's description, record count, and fields.

Call this before query_dataset or summarize_dataset to learn field names and types.

Inputs: layer_url (required). Returns: name, fields, layer_url, description, +4 more

find_nearby_placesFind nearby public placesA
Read-onlyIdempotent
Inspect

Nearest public places of one type, sorted by straight-line distance.

Categories include libraries, public schools, parks, greenways, fire and police stations, post offices, pharmacies, grocery stores, medical facilities, child care, EV chargers, park-and-ride lots, light rail stations, bus stops, YMCAs, places of worship, and public Wi-Fi. Bus stop IDs can be used with CATS real-time arrival services.

Inputs: category (required), location (required), limit, radius_miles. Returns: notes, places, category, location, +2 more

get_311_requests_near311 service requests near a locationA
Read-onlyIdempotent
Inspect

311 service requests reported near a location, with counts by request type.

Shows whether a problem (pothole, missed pickup, dumping, streetlight) was already reported.

Inputs: location (required), days, limit, radius_miles, +1 more. Returns: Records matching a filter, optionally near a location.

get_address_profileAddress profile: representatives, services, zoningA
Read-onlyIdempotent
Inspect

Everything that serves or governs a location, in one call.

Returns city council district and representative, county commissioner, CMPD patrol division, first-due fire station, trash/recycling/yard-waste days, zoning, FEMA flood zone, historic district, ZIP code, and tax parcel ID.

Inputs: location (required). Returns: Public services, representatives, and regulations that apply at a location.

get_bus_routeCATS bus route frequenciesA
Read-onlyIdempotent
Inspect

CATS bus routes with scheduled frequency (minutes between buses) by time of day.

For stops near an address use find_nearby_places with category 'bus_stop'.

Inputs: route. Returns: Records matching a filter, optionally near a location.

get_capital_projects_nearCity capital projects near a locationB
Read-onlyIdempotent
Inspect

City capital projects (roads, sidewalks, parks, facilities) near a location.

Includes phase, schedule, budget, and the project manager's contact information.

Inputs: location (required), limit, phase, radius_miles. Returns: Records matching a filter, optionally near a location.

get_city_budgetCity budget by department or fundB
Read-onlyIdempotent
Inspect

City of Charlotte budget totals for a fiscal year by department, fund, or expense type.

Inputs: limit, group_by, department, fiscal_year. Returns: rows, notes, group_by, fiscal_year, +4 more

get_city_salary_statsCity employee pay statisticsA
Read-onlyIdempotent
Inspect

Pay statistics for City of Charlotte employees: head count and annual pay rates.

Aggregated by department or job title for one quarter (the latest by default); individual employees are not listed.

Inputs: year, limit, quarter, group_by, +2 more. Returns: rows, year, notes, filters, +3 more

get_code_enforcement_casesCode enforcement cases near an addressB
Read-onlyIdempotent
Inspect

Housing and neighborhood code enforcement cases at or near an address.

Covers nuisance (junk, overgrown lots), minimum housing, zoning, graffiti, and parking violations. Useful for renters, buyers, and neighbors.

Inputs: location (required), days, limit, status, +2 more. Returns: Records matching a filter, optionally near a location.

get_crime_nearPolice incidents near a locationA
Read-onlyIdempotent
Inspect

CMPD police incident reports near a location: counts by offense and the latest reports.

Answers "is there much crime near 123 Main St?" or "any car break-ins near me lately?".

Inputs: location (required), days, limit, offense, +1 more. Returns: Records matching a filter, optionally near a location.

get_pending_rezoningsPending rezoning petitionsB
Read-onlyIdempotent
Inspect

Rezoning petitions currently in process, citywide or near a location.

Shows the petitioner, existing and requested zoning, acreage, and a link to the petition.

Inputs: near, limit, radius_miles. Returns: Records matching a filter, optionally near a location.

get_street_closuresStreet closures and detoursA
Read-onlyIdempotent
Inspect

Street and lane closures and detours published by CDOT, citywide or near a location.

Inputs: near, limit, active_only, radius_miles. Returns: Records matching a filter, optionally near a location.

get_traffic_crashes_nearTraffic crashes near a locationA
Read-onlyIdempotent
Inspect

Reported traffic crashes near a location, with counts by severity.

Useful for "is this intersection dangerous?" or traffic-calming requests.

Inputs: location (required), days, limit, radius_miles. Returns: Records matching a filter, optionally near a location.

get_trash_and_recycling_scheduleTrash and recycling scheduleB
Read-onlyIdempotent
Inspect

Garbage, recycling, and yard-waste collection days for an address in Charlotte.

Inputs: location (required). Returns: notes, location, schedule

lookup_addressLook up a Mecklenburg County addressB
Read-onlyIdempotent
Inspect

Find an address in Mecklenburg County's official master address list.

Returns matches with coordinates, tax parcel ID, municipality, and ZIP code.

Inputs: address (required), max_results. Returns: note, query, candidates

query_datasetQuery any datasetB
Read-onlyIdempotent
Inspect

Fetch records from any portal dataset layer with a SQL filter, optionally near a place.

Inputs: layer_url (required), near, limit, where, +3 more. Returns: Records matching a filter, optionally near a location.

search_datasetsSearch Charlotte open datasetsA
Read-onlyIdempotent
Inspect

Search the Charlotte Open Data Portal catalog (about 380 datasets) by keyword.

Use this to find data the purpose-built tools don't cover. Pass a result's layer_url to describe_dataset, query_dataset, or summarize_dataset.

Inputs: query (required), limit. Returns: query, datasets, total_matching

summarize_crimeCitywide crime statisticsA
Read-onlyIdempotent
Inspect

Citywide CMPD incident counts grouped by offense, division, year, or neighborhood.

Use group_by='year' with an offense filter for trends, e.g. robberies per year since 2017.

Inputs: year, limit, offense, group_by, +1 more. Returns: Server-side aggregate of a dataset.

summarize_datasetAggregate any datasetA
Read-onlyIdempotent
Inspect

Count or aggregate any portal dataset server-side, grouped by up to three fields.

Answers questions like "how many 311 requests of each type were made in ZIP 28205?". Rows are sorted by the statistic, largest first, in a column named stat_value.

Inputs: group_by (required), layer_url (required), near, limit, +4 more. Returns: Server-side aggregate of a dataset.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.7/5.0

    Scored across 19 tools

    Disambiguation5/5

    Each tool targets a distinct civic data domain: crashes, crime, 311, code enforcement, closures, rezonings, budget, salaries, and open data. Even the three crime-related tools are cleanly separated by location vs. citywide aggregation.

    Naming Consistency4/5

    Most tools follow a clear verb_noun pattern like get_crime_near, summarize_crime, and describe_dataset. Minor deviations like lookup_address, find_nearby_places, and search_datasets break the strict 'get_' pattern, but the naming is still predictable and readable.

    Tool Count4/5

    Nineteen tools is slightly above the typical sweet spot, but the server covers a broad municipal data platform spanning many domains. Each tool has a clear role, and the dataset meta-tools justify the extra count by enabling open-ended queries beyond the purpose-built tools.

    Completeness5/5

    The tool surface is unusually complete: address lookup, property/government profiles, safety, traffic, 311, code enforcement, planning, transportation, budget, and salary data are all covered. The generic dataset search/describe/query/summarize tools eliminate dead ends by allowing access to the full open data catalog.

    Maintenance

    ActivityNo data
    ResponsivenessOutdated data

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      B
      maintenance
      Provides access to Cincinnati open data via the Socrata SODA API, allowing users to query datasets using natural language or direct tool calls.
      150 npm
      MIT
    • A
      license
      Not graded
      quality
      C
      maintenance
      Enables searching and querying Durham, North Carolina open geospatial data from ArcGIS Feature/Map Services, including parcels, zoning, and public safety datasets via natural language or direct tool calls.
      64 npm
      MIT