Glendale GIS MCP Server
This server lets you look up mapped hazards and city GIS data for Glendale, California by address or coordinates.
Hazard lookups: Get wildfire, flood, seismic (fault rupture, liquefaction, landslide), dam inundation, and debris-flow status at any Glendale location.
Nearest resources: Find nearby fire stations, police, hospitals, schools, libraries, parks, and bus stops with distances.
Dataset queries: Search any of 21 datasets by attribute filters, location, or bounding box, with paging and optional geometry.
Data discovery: List available datasets, describe their fields and coded values, and read guides about the data and real-time sources.
Geocoding: Validate and disambiguate Glendale street addresses into coordinates.
Transparent results: Each answer cites its source, provides feature references, and distinguishes regulatory maps from live conditions—no real-time data is included.
Provides access to Glendale, California hazard and city GIS data sourced from public ArcGIS services, including mapped hazards at an address, nearby community resources, zoning, parcels, and other city layers.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Glendale GIS MCP Server what hazards are at 613 E Broadway?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Glendale GIS MCP Server
An MCP server that gives hackathon teams, and their AI agents, easy access to the hazard maps and city GIS data for Glendale, California: which mapped hazards apply at an address, what community resources are nearby, and the City of Glendale layers behind them.
It was built for Jewel City Hacks 5, for teams working on emergency preparedness: helping residents understand the hazards where they live and how to prepare.
Jewel City Hacks 5
When | Saturday, September 26, 2026, 8:00 AM – 8:00 PM |
Where | Glendale Community College, Math Discovery Center, 1500 N Verdugo Rd, Glendale, CA 91208 |
Theme | "Think Globally, Build Locally": create technology that makes cities smarter, more connected, and more inclusive |
Hosts | Hacker Fund and Glendale Tech, with the City of Glendale Economic Development Division and the GCC Computer Science Club |
Teams | 1–4 people; students, professionals and residents welcome. High school students must be 14 or older, and anyone under 18 needs parental consent. |
Register |
Schedule: check-in 8:00 AM · project submission closes 6:00 PM · demo expo 6:15 PM (5-minute presentations) · closing ceremony and winners 7:15 PM · check-out by 8:00 PM.
Judging: technical execution 40%, social/local impact 20%, presentation 15%, feasibility 15%, creativity 10%. Prizes: Grand Champion, Best Hardware Hack, Best Software Hack, and Top 5 Hackathon Hackers.
Event details are from hacker.fund/jewelcityhacks and the Luma page as of September 18, 2026. Check there for changes. Questions about the event: team [at] hacker.fund.
Related MCP server: mcp-arcgis-tucson
What this server gives you
It finds, queries and normalizes authoritative data. It doesn't interpret risk, rank hazards or give advice; that's what teams build.
Ask | Tool |
Which hazards apply at an address or point? |
|
Nearest fire stations, police, hospitals, schools, libraries, parks, bus stops? |
|
Zoning or parcel at a location, schools of a type, bus stops on a route, any layer by attribute or area? |
|
What data is there, and what do the fields mean? |
|
Is this a valid Glendale address? |
|
What can this server do? Where's live emergency information? |
|
Every tool takes a location as a street address ({"address": "613 E Broadway"}) or coordinates ({"lat": 34.1466, "lon": -118.2483}). Results cite their source, and each feature carries a ref for fetching the full live record.
Data (21 datasets):
Hazards: CAL FIRE Fire Hazard Severity Zones (2025), California Geological Survey fault rupture, liquefaction and landslide zones, FEMA flood zones, DWR dam inundation areas, and USGS post-fire debris-flow assessments.
City of Glendale: city boundary, fire stations and districts, police station, hospitals, schools, libraries, parks, Beeline bus stops, ZIP codes, neighborhood zones, zoning, streets, and parcels (queried live).
Most data comes from an offline snapshot of the public ArcGIS services (about 35 MB), so queries are fast and don't load the city's servers.
Not included: real-time information (active fires, evacuation orders, weather warnings, earthquakes). See Real-Time Emergency Information for the official sources and feeds to use instead.
Try it
You need Python 3.10 or later. On first run the server downloads the data snapshot (about 6 MB) from this repository's releases, checks its SHA-256, and caches it. Install it one of two ways, then connect your assistant below.
Option A: uv (recommended)
uv runs the server straight from GitHub, with nothing to install permanently:
uvx --from git+https://github.com/HackerFund/GlendaleGisMcp glendale-gis-mcp --fetch-snapshotThat downloads and verifies the data, then exits, so your assistant doesn't wait for it later. The command to give your assistant is:
uvx --from git+https://github.com/HackerFund/GlendaleGisMcp glendale-gis-mcpOption B: pip
python3 -m venv ~/glendale-gis-mcp
source ~/glendale-gis-mcp/bin/activate # Windows: %USERPROFILE%\glendale-gis-mcp\Scripts\activate
pip install git+https://github.com/HackerFund/GlendaleGisMcp
glendale-gis-mcp --fetch-snapshot # download and verify the dataThe command to give your assistant is the full path ~/glendale-gis-mcp/bin/glendale-gis-mcp (Windows: %USERPROFILE%\glendale-gis-mcp\Scripts\glendale-gis-mcp.exe). To update later, force a reinstall, since the version number doesn't change between commits:
pip install --force-reinstall --no-deps git+https://github.com/HackerFund/GlendaleGisMcpOption C: clone this repository
Run the code from a checkout, which is also what you want if you plan to change it:
git clone https://github.com/HackerFund/GlendaleGisMcp.git
cd GlendaleGisMcp
python3 -m venv .venv && source .venv/bin/activate
pip install -e . # add ".[dev]" for the tests
glendale-gis-mcp --fetch-snapshot # download and verify the dataThe command to give your assistant is the absolute path printed by echo $PWD/.venv/bin/glendale-gis-mcp.
To pull updates later: git pull, then pip install -e . again if the dependencies changed. To build your own snapshot instead of using the published one (about 1.5 minutes, and it queries the live agency servers):
python scripts/build_snapshot.py
export GLENDALE_GIS_SNAPSHOT_PATH=$PWD/snapshotConnect your AI assistant
The examples use the uv command. With pip or a clone, swap it for the full path to glendale-gis-mcp and drop the --from … arguments.
Claude Code
claude mcp add glendale-gis -- uvx --from git+https://github.com/HackerFund/GlendaleGisMcp glendale-gis-mcpCheck it with claude mcp list. (With pip: claude mcp add glendale-gis -- ~/glendale-gis-mcp/bin/glendale-gis-mcp.)
Claude Desktop
Open Settings → Developer → Edit Config, add the entry below, then quit Claude Desktop with ⌘Q and reopen it. Use the full path to uvx (which uvx), because Claude Desktop doesn't see your shell's PATH:
{
"mcpServers": {
"glendale-gis": {
"command": "/full/path/to/uvx",
"args": ["--from", "git+https://github.com/HackerFund/GlendaleGisMcp", "glendale-gis-mcp"]
}
}
}With pip, use "command": "/Users/you/glendale-gis-mcp/bin/glendale-gis-mcp" and no args.
uv caches the code it downloaded, so a restart alone may keep an older commit. To always check GitHub, add "--refresh" as the first item in args. To pin a version for the event, put a tag or commit on the URL: git+https://github.com/HackerFund/GlendaleGisMcp@<tag>.
Gemini CLI
Add this to ~/.gemini/settings.json (or .gemini/settings.json in a project), then restart the CLI:
{
"mcpServers": {
"glendale-gis": {
"command": "uvx",
"args": ["--from", "git+https://github.com/HackerFund/GlendaleGisMcp", "glendale-gis-mcp"]
}
}
}The Gemini CLI also has gemini mcp add <name> <command> [args...]; put -- before the command so its own flags aren't confused with uv's. See Google's MCP documentation for the current syntax. Use /mcp inside the CLI to list connected servers.
ChatGPT
ChatGPT can't run this server today. Its custom connectors only reach remote MCP servers over HTTPS, so a local install like the above isn't an option, and its connectors authenticate with OAuth or no authentication, per OpenAI's documentation. Our hosted server uses a shared key sent as an HTTP header, which ChatGPT doesn't send.
Custom connectors also need Developer Mode (Settings → Apps → Advanced) and a paid plan.
If you want to use ChatGPT, the options are:
Use Claude, the Gemini CLI or another MCP client for the data, and ChatGPT for the rest of your build.
Ask the organizers for an open (no-key) endpoint for your team.
Call the
corepackage directly from your own Python code (see Development) and pass the results to ChatGPT.
Other clients
Cursor, VS Code, Zed, Windsurf and most other MCP clients take the same command and arguments; check their docs for where the config lives. To poke at the server by hand:
npx @modelcontextprotocol/inspector uvx --from git+https://github.com/HackerFund/GlendaleGisMcp glendale-gis-mcpHosted server (Google Cloud Run)
Teams can use the hosted server instead of installing anything. It speaks streamable HTTP and needs the shared hackathon key on every request. The organizers hand out the key through the event channel.
MCP endpoint:
https://glendale-gis-mcp-1053589358088.us-west2.run.app/mcpKey:
Authorization: Bearer <key>header (never in the URL)Health check (no key needed):
curl https://glendale-gis-mcp-1053589358088.us-west2.run.app/healthRate limit: 120 requests a minute per key. Over that you get
429with aRetry-Afterheader.
A missing or wrong key returns 401. If the hosted server is down, the local options above give you the same tools and data. Deployment steps are in docs/deploy.md.
Claude Code:
claude mcp add --transport http glendale-gis https://glendale-gis-mcp-1053589358088.us-west2.run.app/mcp \
--header "Authorization: Bearer <key>"Gemini CLI (~/.gemini/settings.json):
{
"mcpServers": {
"glendale-gis": {
"httpUrl": "https://glendale-gis-mcp-1053589358088.us-west2.run.app/mcp",
"headers": { "Authorization": "Bearer <key>" }
}
}
}Check Google's MCP documentation for the current field names.
Claude Desktop and ChatGPT can't send a custom header to a remote MCP server, so they can't use the hosted endpoint with a key. Use a local install for those.
Keep the key out of anything you publish: no screenshots, no commits, no Devpost submissions.
Try a question
Ask, for example: "Is 613 E Broadway in Glendale in a flood zone?" or "What's the nearest fire station to 1000 W Glenoaks Blvd?" Ask "What can the Glendale GIS server do?" for a tour.
Development
Clone it as in Option C, install with pip install -e ".[dev]", then:
pytest # 339 tests; none of them touch the network
ruff check . && ruff format .AGENTS.md holds the project's rules and design decisions, and Plans/ has the research notes and the phased plan. The core package has no MCP imports, so you can use it as a plain Python library:
from glendale_gis.core.snapshot import Snapshot
from glendale_gis.core import hazards
snap = Snapshot.load("snapshot") # or the downloaded copy in your cache
result = hazards.hazards_at_location(snap, hazards.locate(snap, 34.1466, -118.2483))
print(result.wildfire.status, result.flood.matches[0].attributes["FLD_ZONE"])Docs
Reading the Snapshot Data: GeoJSON, fields, each layer, and how to read each hazard
Real-Time Emergency Information: official alerts and live data feeds
Inside a connected client, call
read_guide("about")for the full tour.
Please read
The hazard layers are regulatory maps, not site-specific assessments. Being outside a mapped zone doesn't mean there's no hazard. For example, CAL FIRE "NonWildland" means unzoned, not safe from wildfire.
City of Glendale GIS data is for general information and is not a substitute for legal descriptions or surveys.
This server has no real-time data. In an emergency, follow official alerts.
License
Copyright (C) 2026 Ryan Gates
This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. It is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See LICENSE for details.
The data comes from public agency services and keeps their terms and disclaimers.
Available Tools
12 toolsdam_inundationARead-onlyIdempotent
DWR dam inundation areas at a location: where water could flood if a dam failed. One dam can have several features (failure scenarios or structures); all are returned. HazardCl rates the consequences of a failure, not its likelihood. Federally owned dams are not included. Not an evacuation map.
Inputs: location (required). Returns: _meta, notes, title, reason, +8 more
debris_flowARead-onlyIdempotent
USGS post-fire debris-flow assessments at a location. USGS only assesses recently burned areas, and none currently covers Glendale, so this usually returns not_in_zone; that does not mean there is no debris-flow risk.
Inputs: location (required). Returns: _meta, notes, title, reason, +8 more
describe_datasetARead-onlyIdempotent
Describe one dataset: every field with its type, meaning and coded values (inferred meanings are flagged), record count, freshness (fetched_at, source_last_edit), the area the snapshot covers, the source URL and disclaimer. Read this before filtering with query_dataset, and to interpret attributes returned by other tools (e.g. FEMA FLD_ZONE codes, CAL FIRE FHSZ codes).
Inputs: dataset (required). Returns: id, clip, docs, _meta, +17 more
flood_zoneARead-onlyIdempotent
FEMA flood zone at a location (National Flood Hazard Layer), with the nearest zone of each FLD_ZONE class. Every FEMA zone counts as in_zone, including X (minimal or 0.2%-annual-chance hazard); read FLD_ZONE, ZONE_SUBTY and SFHA_TF (T = Special Flood Hazard Area). Zone D means not studied, not safe.
Inputs: location (required). Returns: _meta, notes, title, reason, +8 more
geocode_addressARead-onlyIdempotent
Find a Glendale street address with the City of Glendale geocoder. Returns the matched address, coordinates and score, or several candidates when the address is ambiguous (e.g. '613 Broadway' matches both E and W Broadway). Addresses outside the city limits (the geocoder also knows Burbank and unincorporated La Crescenta) return an error. Unit numbers are ignored and place names aren't supported. Other tools accept an address directly, so call this only to check or disambiguate one.
Inputs: address (required). Returns: _meta, notes, status, location, +1 more
hazards_at_locationARead-onlyIdempotent
All mapped hazards at one location in one call: wildfire (CAL FIRE), flood (FEMA), fault rupture, liquefaction and earthquake-induced landslide (CGS), dam inundation (DWR) and post-fire debris flow (USGS). Each has a status (in_zone, not_in_zone or unavailable), the zones containing the point, the nearest zone with its distance, and notes on how to read it. These are regulatory maps of mapped hazard, not current conditions or a safety assessment; for active fires, evacuations and alerts call read_guide("real_time_sources"). Use the single-hazard tools when only one is needed.
Inputs: location (required). Returns: fault, flood, location, wildfire, +4 more
list_datasetsARead-onlyIdempotent
List every dataset this server can query: id, source agency, category (hazard, resource or reference), whether it comes from the offline snapshot or live, and its feature count. Start here to find dataset ids for describe_dataset and query_dataset. For hazards or resources at a location, the dedicated tools (hazards_at_location, nearest_resources) are simpler.
Inputs: no named parameters. Returns: notes, datasets, snapshot_built_at
nearest_resourcesARead-onlyIdempotent
Nearest community resources to a location: fire stations, police station, hospitals, schools, libraries, parks and Beeline bus stops, with name, address and distance. Distances are straight-line meters, not travel distance or time, and only resources inside Glendale are included (neighboring cities' facilities may be closer). For other filters (e.g. only high schools) use query_dataset.
Inputs: location (required), kinds, limit. Returns: notes, groups, location
query_datasetARead-onlyIdempotent
Search one dataset with attribute filters, a location (near) or a bounding box, with paging. Use it for questions the dedicated tools don't cover: zoning or the parcel at a location, schools of a type, bus stops on a route, all dams affecting an area. Results carry total, next_offset for the next page, and a ref for each feature. Geometry is off by default to save context. Parcels are queried live from the city; everything else comes from the offline snapshot.
Inputs: dataset (required), bbox, near, limit, +4 more. Returns: _meta, notes, total, offset, +4 more
read_guideARead-onlyIdempotent
Read one of the server's guides (Markdown). Call "about" first if you're unsure what this server can do. Call "real_time_sources" whenever someone asks about current conditions (active fires, evacuations, warnings, earthquakes, outages): this server has no real-time data, and the guide lists the official sources to send them to.
Inputs: topic (required). Returns: title, topic, content, resource_uri
seismic_zonesARead-onlyIdempotent
California Geological Survey earthquake zones at a location: Alquist-Priolo fault rupture, liquefaction and earthquake-induced landslide. Being inside a zone is the hazard signal; the attributes only describe the map (quadrangle, release date, report links). Outside a zone does not mean no earthquake hazard: all of Glendale can shake.
Inputs: location (required). Returns: fault, location, landslide, liquefaction
wildfire_zoneARead-onlyIdempotent
CAL FIRE Fire Hazard Severity Zone at a location (2025 map), with the nearest zone of each class (Very High, High, Moderate, NonWildland). The zones rate the long-term hazard of the landscape, not the risk to a building. NonWildland is unzoned, not safe from wildfire. Not an evacuation map; for active fires call read_guide("real_time_sources").
Inputs: location (required). Returns: _meta, notes, title, reason, +8 more
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 12 tools
Every tool has a clearly distinct purpose: listing datasets, reading guides, describing schemas, querying data, geocoding, and specific hazard lookups per type (wildfire, flood, seismic, dam, debris) plus resource search. The aggregate hazards_at_location is explicitly differentiated from single-hazard tools.
All tool names use lowercase snake_case with a consistent style, mostly verb_noun (list_datasets, read_guide, describe_dataset, query_dataset, geocode_address) or resource nouns (wildfire_zone, flood_zone, seismic_zones). The pattern is uniform and predictable.
12 tools is well within the ideal 3-15 range. Each tool covers a distinct aspect of the GIS domain (hazards, resources, querying, geocoding) without redundancy, and the count matches the server's comprehensive scope.
The tool surface provides complete coverage for the stated purpose: listing available datasets, describing schema, flexible querying (attribute, location, bbox), geocoding, all major hazard types, and community resources. No obvious gaps; read-only GIS queries are fully supported.
Maintenance
Related MCP Connectors
US weather & geo for AI agents: forecasts, alerts, earthquakes, elevation, geocoding. No keys.
US weather & geo for AI agents: forecasts, alerts, earthquakes, elevation, geocoding. No keys.
Flood, wildfire, soils, terrain, power grid, utilities and parcels for any US location, sourced.
Flood, wildfire, earthquake, and coastal-storm hazard lookup for a US property.
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceEnables AI agents to search and query Los Angeles County open geospatial datasets (parcels, parks, etc.) via ArcGIS Feature Services.76 npmMIT
- AlicenseNot gradedqualityCmaintenanceEnables users to search and query City of Tucson GIS open geospatial datasets (zoning, parcels, public safety, transport, city services) through natural language.84 npmMIT
- AlicenseNot gradedqualityCmaintenanceEnables searching and querying City of Henderson open geospatial datasets (parcels, zoning, public works) via natural language or direct tool calls.56 npmMIT
- AlicenseNot gradedqualityCmaintenanceEnables search and query of City of Peoria, Arizona open geospatial datasets (parcels, zoning, public works) via ArcGIS feature services, allowing AI agents to access GIS data through natural language.85 npmMIT