ontario-data-mcp — An MCP server for Ontario's open data

What happens behind a question

Behind every question is a search, a download, a schema check, a SQL query. Expand any card to see exactly what happened.

Time series

How fast is EV adoption growing across Ontario?

Caches 17 quarterly CSVs of EV registrations. Runs time series with window functions. Province-wide EVs grew 3.2× in three years. Northern Ontario shows fastest growth rate.

Source: EV Registrations by FSA

Under the hood

search_datasets → b5696d0c

download_resource → 17 CSVs

query_cached

SELECT FSA, ROUND(Total_EV * 1.0 / LAG(Total_EV) OVER ( PARTITION BY FSA ORDER BY quarter ), 2) AS qoq_growth FROM ev_registrations ORDER BY qoq_growth DESC

Cross-dataset

Is there a link between ambulance and ER wait times?

Downloads ambulance response data and hospital ER waits. Joins on health region. Regions with slow response also show 22% longer ER waits.

Sources: Ambulance Times, Wait Times

Under the hood

search_datasets × 2

download_resource × 2

query_cached

SELECT amb.region, amb.avg_response, er.avg_wait FROM ambulance_times amb JOIN er_waits er ON amb.region = er.region ORDER BY avg_response DESC

170 datasets

How have EQAO math scores changed over a decade?

Pulls 14 XLSX files of school demographics. Pivots Grade 6 math by board. Provincial average dropped 8 points 2016–2019, partially recovered post-pandemic.

Source: School Demographics

Under the hood

search_datasets → d85f68c5

list_resources → 14 XLSX

download_resource

query_cached

SELECT Board_Name, School_Year, AVG(Pct_Grade_6_Math) AS pct FROM school_demographics GROUP BY Board_Name, School_Year ORDER BY Board_Name, School_Year

Geospatial

Where are Ontario's child care deserts?

Caches child care facilities, loads census boundaries. Calculates spots per child by ward. 6 wards in Scarborough have <1 spot per 10 children.

Source: Licensed Child Care Facilities

Under the hood

search_datasets → 7efd8b4b

download_resource

load_geodata → ward boundaries

spatial_query

SELECT w.ward_name, SUM(c.capacity) AS spots FROM wards w LEFT JOIN childcare c ON ST_Contains(w.geom, ST_Point(c.lng, c.lat)) GROUP BY w.ward_name

Investigation

Which facilities are Ontario's biggest air polluters?

Downloads 77 Environmental Compliance resources. Profiles data, queries by facility. Top 10 emitters account for 34% of total NOx.

Source: Environmental Compliance Reports

Under the hood

search_datasets → 13cab7d0

list_resources → 77 resources

download_resource

profile_data

query_cached

SELECT Facility_Owner, SUM(Exceedance) AS total FROM air_emissions WHERE Contaminant = 'NOx' GROUP BY Facility_Owner ORDER BY total DESC

Cross-dataset

How does Great Lakes water quality compare to inland?

Caches lake intakes (1976–2024) and provincial monitoring (24 resources). Joins on parameter. Inland phosphorus runs 2.4× higher than lake intakes.

Sources: Lake Intakes, Provincial Monitoring

Under the hood

search_datasets × 2

download_resource × 2

query_cached

SELECT l.yr, l.avg_p AS lake_p, i.avg_p AS inland_p, ROUND(i.avg_p / l.avg_p, 1) FROM lakes l JOIN inland i ON l.yr = i.yr

← Scroll for more →

24 tools under the hood

Expand any category to see the individual tools.

Discovery 6 tools

Search by keyword, topic, ministry, or geographic area.

search_datasets, list_organizations, list_topics, get_popular_datasets, search_by_location, find_related_datasets

Metadata 4 tools

Inspect schemas, resources, and compare datasets side by side.

get_dataset_info, list_resources, get_resource_schema, compare_datasets

Retrieval 4 tools

Download and cache datasets locally in DuckDB for fast, offline access.

download_resource, cache_info, cache_manage, refresh_cache

Querying 4 tools

Run SQL against remote or cached data. Preview rows or run custom queries.

query_resource, sql_query, query_cached, preview_data

Data Quality 3 tools

Freshness checks, completeness analysis, and statistical profiles.

check_data_quality, check_freshness, profile_data

Geospatial 3 tools

Load shapefiles, KML, and GeoJSON into DuckDB. Run spatial queries.

load_geodata, spatial_query, list_geo_datasets

Guided workflows

Built-in prompts that run multi-step tool sequences. Each one searches the live catalogue, checks your cache, and builds context before the LLM responds.

explore_topic

Explore housing data in Ontario

Searches the catalogue for "housing," returns the top results across ministries, inspects metadata on the most relevant datasets, and recommends which to download. Finds 40+ datasets spanning rent tribunals, social housing waitlists, and building permits.

What the prompt triggers

search_datasets → "housing" (live catalogue search, top 5 shown)

get_dataset_info → inspects interesting results

list_resources → checks formats, row counts

Recommends downloads, suggests questions you can answer

Usage: explore_topic(topic="housing")

data_investigation

Investigate the licensed child care facilities dataset

Pulls full metadata, downloads the primary CSV, runs schema inspection, data quality checks, and statistical profiling. Surfaces 12% null rates in postal codes, flags capacity outliers, and suggests spatial joins with census boundaries.

What the prompt triggers

get_dataset_info → full metadata

list_resources → finds primary CSV

get_resource_schema → column types and names

download_resource → caches in DuckDB

check_data_quality → nulls, duplicates, freshness

profile_data → distributions, outliers, patterns

query_cached → follow-up SQL for time series or correlations

Usage: data_investigation(dataset_id="licensed-child-care-facilities-in-ontario")

compare_data

Compare ambulance response times with ER wait times

Downloads both datasets, profiles each independently, then runs cross-dataset analysis. Finds shared region columns, confirms joinability, and shows regions with slow ambulance response also have 22% longer ER waits.

What the prompt triggers

compare_datasets → metadata comparison

download_resource × 2 → caches both datasets

profile_data × 2 → structure, size, quality for each

query_cached → DuckDB SQL to find shared columns and join

Summarizes complementarity and join potential

Usage: compare_data(dataset_ids="ambulance-response-times, ontario-wait-times")

Prompts 3 prompts

Context-aware guided workflows with live catalogue data.

explore_topic — guided exploration of any topic
data_investigation — deep dive: schema, quality, stats
compare_data — side-by-side dataset analysis

Resources 4 resources

Read-only context for the LLM: cache status, metadata, SQL tips.

ontario://cache/index — cached datasets with freshness
ontario://dataset/{id} — full dataset metadata
ontario://portal/stats — catalogue overview
ontario://guides/duckdb-sql — SQL tips for Ontario data

Go deeper with follow-ups

One question leads to another. The tool remembers context.

Cross-dataset

Do areas with more EVs have better air quality?

Builds on cached EV data. Joins with air quality readings by postal area. GTA high-adoption zones show 12% lower NOx.

Sources: EV Registrations, Air Quality

Under the hood

Uses EV data already cached, then:

search_datasets → air quality

download_resource

query_cached

SELECT ev.FSA, ev.Total_EV, AVG(aq.NO2) AS avg_no2 FROM ev_registrations ev JOIN air_quality aq ON LEFT(aq.postal, 3) = ev.FSA GROUP BY ev.FSA, ev.Total_EV

Geospatial

Map child care deserts in detail.

Builds on cached child care data. Loads census boundaries, runs spatial join. 6 wards have <1 spot per 10 children.

Source: Licensed Child Care Facilities

Under the hood

Uses child care data already cached, then:

load_geodata → census boundaries

spatial_query

SELECT b.area, SUM(c.capacity) / b.pop_u5 AS spots_per_child FROM census_boundaries b LEFT JOIN childcare c ON ST_Contains(b.geom, ST_Point(c.lng, c.lat)) GROUP BY b.area

Guided

I have data cached but don't know where to start.

The data_investigation prompt auto-runs schema checks, quality analysis, and suggests queries. Surfaces nulls, outliers, and patterns.

Resource: ontario://cache/index

Under the hood

data_investigation prompt triggers:

ontario://cache/index

get_resource_schema

check_data_quality

profile_data

query_cached → suggested queries

Freshness

Are my cached datasets still current?

Checks staleness across your cache. Refreshes only what's outdated. School data updates monthly; fire perimeters daily.

Resource: ontario://cache/index

Under the hood

ontario://cache/index

// Returns: ev_registrations 3d ago fresh school_data 38d ago stale childcare 12d ago fresh

check_freshness

refresh_cache → stale only

Discovery

What education data does Ontario publish?

Searches 170+ datasets, finds related sources, compares schemas. Recommends best fit for your analysis.

Browse: 170+ education datasets

Under the hood

explore_topic → "education"

search_datasets → 170 results

find_related_datasets

compare_datasets

// Comparison: School Demographics 14 res XLSX Board Expenditures 8 res CSV Public School List 4 res TXT

Warehouse

Can I build a local data warehouse?

Cache 10+ datasets into DuckDB. Run window functions, CTEs, pivots across domains. Impossible through the remote API.

Guide: ontario://guides/duckdb-sql

Under the hood

download_resource × 10+

ontario://guides/duckdb-sql

query_cached

WITH regional AS ( SELECT region, AVG(amb.response), AVG(aq.PM25), AVG(sch.math_pct) FROM ambulance amb JOIN air_quality aq ... JOIN schools sch ... )

← Scroll for more →

Get started

Works with any MCP-compatible client. No API key required.

Add to claude_desktop_config.json

{
  "mcpServers": {
    "ontario-data": {
      "command": "uvx",
      "args": ["ontario-data-mcp"]
    }
  }
}

Or clone from source: git clone https://github.com/sprine/ontario-data-mcp && cd ontario-data-mcp && uv sync

An MCP server for Ontario's open data.

What happens behind a question

How it works

24 tools under the hood

Guided workflows

Go deeper with follow-ups

Get started