Worcester GIS
БесплатноНе проверенEnables querying and exploring the City of Worcester's open data portal (ArcGIS Hub) through natural language, including dataset search, type filtering, schema
Описание
Enables querying and exploring the City of Worcester's open data portal (ArcGIS Hub) through natural language, including dataset search, type filtering, schema inspection, spatial point-in-polygon queries, and geocoding.
README
License: MIT Python 3.11+ MCP Compatible
San Diego City GIS MCP — a City of San Diego fork of OpenContext (forked from the Worcester GIS fork). It serves the City's ArcGIS Server REST services directory (webmaps.sandiego.gov/arcgis/rest/services, ArcGIS Enterprise 11.5) through the built-in arcgis plugin.
It is a sibling to the San Diego regional (SANDAG/SanGIS) and Worcester servers: identical tool names and signatures, so it composes with them at the MCP client with zero new orchestration. The regional server covers county-wide layers; this one covers City-authored layers (MHPA, Base Zones, Community Plan Land Use, and ~700 more).
How discovery works (no Hub here)
webmaps.sandiego.gov is a bare services directory — there is no ArcGIS Hub / Open Data catalog in front of it, so the Hub-search discovery used by the Worcester fork does not apply. Instead:
scripts/crawl_catalog.pywalks the directory offline: folders → MapServer/FeatureServer services → each service's/layers?f=json, capturing layer id, name, geometry type, description, extent, andmaxRecordCount.- The result is serialized to
plugins/arcgis/catalog.json— a versioned, diffable deploy artifact (~718 layers from 329 services at last crawl). - The running server loads that manifest at startup — instant, no live crawl, no cold-start penalty.
search_datasetsdoes substring/fuzzy/acronym matching over it. - Services that require an ArcGIS account (HTTP 401/403 or ArcGIS error codes 498/499) are skipped during the crawl and recorded in the manifest's
skippedlist — 15 folders on this host at last crawl (AMPGIS, GetItDone, TED, …).
To refresh the catalog: python scripts/crawl_catalog.py, review the diff, commit, redeploy.
Dataset IDs
A dataset id is the layer's services-directory path:
{folder}/{service}/{MapServer|FeatureServer}/{layerId}
Known-good public layers (all in the anonymous Planning/PLN_LongRangePlanning MapServer):
| Layer | dataset_id | Notes |
|---|---|---|
| Multi-Habitat Planning Area (MHPA) | Planning/PLN_LongRangePlanning/MapServer/7 |
Polygon. Key fields: HABPRES (int, % targeted preservation), INHABPRES, SUBAREA, ACRES |
| Base Zones (official City zoning) | Planning/PLN_LongRangePlanning/MapServer/27 |
Polygon. Zone codes like RS-1-7, CC-3-5 |
| Community Plan Land Use | Planning/PLN_LongRangePlanning/MapServer/24 |
Polygon |
These are curated as featured_datasets in config.yaml (aliases + notes boost search; edits take effect on deploy without re-crawling).
The WGS84 contract
The layers are authored in EPSG:2230 (NAD83 State Plane California Zone VI, US survey feet). Every query this server sends sets inSR=4326 and outSR=4326, so all tools take and return WGS84 lon/lat — same as the sibling servers. Without inSR, WGS84 coordinates would be interpreted as State Plane feet and silently return zero rows.
Pagination is metadata-driven: each layer's maxRecordCount comes from the catalog (it varies by layer), and query_data pages with resultOffset/resultRecordCount when needed.
Connect to the server
Add it as a custom connector in Claude (same steps on Claude.ai and Claude Desktop):
Settings → Connectors (or Customize → Connectors on claude.ai)
Add custom connector
Name it e.g.
San Diego City GISand paste the URL:https://sandiego-city-gis.codeforanchorage.org/mcp
Quick health check from a terminal:
curl -sS -X POST https://sandiego-city-gis.codeforanchorage.org/mcp \
-H 'Content-Type: application/json' \
-d '{"jsonrpc":"2.0","id":1,"method":"ping"}'
# → {"jsonrpc":"2.0","id":1,"result":{"status":"ok"}}
Tools exposed
| Tool | Purpose |
|---|---|
arcgis__search_datasets |
Discover layers by keyword (e.g. "MHPA", "zoning"). Matches names, service/folder names, descriptions, acronyms. Optional type filter: MapServer/FeatureServer or a geometry (Polygon, Point, Polyline) |
arcgis__get_dataset |
Fetch a layer's metadata: geometry type, record cap, extent, layer URL |
arcgis__get_layer_schema |
List a layer's fields (name, type, alias, coded values), optionally filtered by keyword |
arcgis__get_distinct_values |
List the distinct values in a field (with optional like / where) to confirm exact codes |
arcgis__query_data |
Query features (supports where, out_fields, order_by, limit). Output leads with a TOTAL MATCHING count, so "how many X?" needs no paging. Auto-paginates past per-layer record caps |
arcgis__spatial_query_point |
Point-in-polygon: which polygon(s) contain a point — by lon/lat (WGS84) or a street address |
arcgis__geocode_address |
Convert a street address to lon/lat (US Census geocoder, biased to San Diego) |
arcgis__get_aggregations |
Facet counts of the catalog by folder, service, service_type, or geometry_type |
Raw JSON-RPC example:
curl -sS -X POST https://sandiego-city-gis.codeforanchorage.org/mcp \
-H 'Content-Type: application/json' \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/call",
"params":{"name":"arcgis__search_datasets",
"arguments":{"q":"MHPA","limit":5}}}'
Verified end-to-end
The definition-of-done check is a WGS84 point-in-polygon on MHPA at the Tijuana River Valley (32.5539, -117.0846) — a point that returns null on the regional server's County MSCP_CN layer but sits inside a City MHPA preserve:
// arcgis__spatial_query_point
{
"item_id": "Planning/PLN_LongRangePlanning/MapServer/7",
"lon": -117.0846,
"lat": 32.5539,
"out_fields": "HABPRES,INHABPRES,SUBAREA,ACRES"
}
returns the containing preserve polygon:
Record 1:
HABPRES: 100
INHABPRES: Yes
SUBAREA: 113
ACRES: 2734.79
scripts/smoke_prod.py runs this plus 12 more checks (search resolution for "MHPA" and "zoning", schema, TOTAL MATCHING counts, geocode → zoning chain at City Hall) against any deployment:
python scripts/smoke_prod.py # production
python scripts/smoke_prod.py http://localhost:8000/mcp # local
Local development
uv sync # or: pip install -r requirements.txt
python scripts/crawl_catalog.py # (re)build plugins/arcgis/catalog.json
python scripts/local_server.py # serves http://localhost:8000/mcp
python -m pytest tests/ -q # tests
See CLAUDE.md and docs/ for architecture, deployment, and plugin development.
Установка Worcester GIS
У этого сервера нет опубликованного пакета — он собирается из исходников. Открой репозиторий и следуй инструкции в README.
▸ github.com/codeforanchorage/sandiego-city-mcpFAQ
Worcester GIS MCP бесплатный?
Да, Worcester GIS MCP бесплатный — установка в пару кликов через Unyly без оплаты.
Нужен ли API-ключ для Worcester GIS?
Нет, Worcester GIS работает без API-ключей и переменных окружения.
Worcester GIS — hosted или self-hosted?
Self-hosted: сервер запускается локально на твоей машине командой из раздела установки.
Как установить Worcester GIS в Claude Desktop, Claude Code или Cursor?
Открой Worcester GIS на unyly.org, выбери вкладку своего клиента (Claude Desktop, Claude Code, Cursor) и нажми Install — конфиг сгенерируется автоматически, без правки JSON.
Похожие MCP
GitHub
PRs, issues, code search, CI status
автор: GitHubFilesystem
Secure file operations with configurable access controls.
Memory
Knowledge graph-based persistent memory system.
Template MCP Server
A CLI tool to create a new Model Context Protocol server project with TypeScript support, dual transport options, and an extensible structure
автор: mcpdotdirectCompare Worcester GIS with
Не уверен что выбрать?
Найди свой стек за 60 секунд
Автор?
Embed-бейдж для README
Похожее
Все в категории development
