aihw-mcp

MCP server for Australian Institute of Health and Welfare statistics. Plain-English access to long-term mortality (GRIM), regional mortality (MORT), cancer incidence and mortality (ACIM), national health expenditure, youth justice detention, and the public hospitals register — all from a single uvx command.

"How have diabetes deaths changed since 1980?"
"What's the age-standardised mortality rate in the Sydney - Inner West SA3?"
"Show me breast cancer incidence in women aged 50–54 over time."
"How much did Australia spend on public hospitals in 2022-23?"
"How many young people are in detention in NSW vs VIC?"
"List all Principal referral hospitals in Queensland."
"Top 5 causes of death in 2023."

Sister to abs-mcp (Australian Bureau of Statistics), rba-mcp (Reserve Bank of Australia), ato-mcp (Australian Taxation Office), and au-weather-mcp (Australian weather). The five together cover the macro / regulator / tax / health / climate layer of Australian official data.

---

Install

# Run on demand via uvx (recommended)
uvx --upgrade aihw-mcp

# Or install permanently
pip install aihw-mcp

Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "aihw": { "command": "uvx", "args": ["--upgrade", "aihw-mcp"] }
  }
}

Why --upgrade? uvx aihw-mcp (without the flag) uses whatever wheel is cached and never adopts new PyPI releases on its own. --upgrade makes uvx check PyPI on each launch and pull a newer release if one exists. To verify which version is currently serving you, look at the server_version field on any DataResponse.

Claude Code / Cursor

claude mcp add aihw --command uvx --args -- --upgrade aihw-mcp

Auto-updating data

Beyond the wheel-level --upgrade, the server has a second auto-update path inside the data layer: when AIHW refreshes GRIM with another year of deaths data or publishes a new MORT release, aihw-mcp resolves the new resource URL via data.gov.au's CKAN API at fetch time and uses the freshest match. Hard-coded YAML URLs are the safe fallback if discovery fails. You do not need to wait for a new wheel release to get new yearly data — just delete ~/.aihw-mcp/cache.db to force a refresh, or wait for the 7-day TTL to expire.

---

What it exposes

Six tools, all plain-English in, structured out:

Tool	Purpose
search_datasets	Fuzzy-search the curated catalog by keyword
describe_dataset	List a dataset's filterable dimensions and returnable measures
get_data	Query with filters, measures, period range, output format
latest	Last observation per measure (shortcut)
top_n	Rank rows by a measure, return top (or bottom) N
list_curated	Enumerate the curated dataset IDs

Every response is the same shape — dataset_id, dataset_name, query, period, unit, row_count, records, aihw_url, attribution, server_version — across every curated dataset.

Time-series datasets (GRIM, MORT, ACIM, Health Expenditure, Youth Justice) accept start_period and end_period on get_data — e.g. start_period="2000", end_period="2010" narrows GRIM to that decade. latest() returns the most-recent observation per measure, sorted by the dataset's declared period dimension (not by source row order). Error messages include fuzzy "did you mean?" suggestions when you typo a filter or measure name.

---

Curated datasets (6 in v0.1)

ID	What it is	Period	Coverage
GRIM_DEATHS	National long-term mortality: deaths × cause × year × sex × age band	1907 → present	~370k rows, 3 measures
MORT_GEOGRAPHY	Regional mortality: deaths + premature/avoidable deaths × State / SA3 / SA4 / PHN	2019 → present	~15k rows, 15 measures
CANCER_INCIDENCE_MORTALITY	Cancer incidence + mortality counts × year × sex × type × 5-year age band	1968 → present	~9k rows, 19 age columns
HEALTH_EXPENDITURE	Real expenditure by financial year × state × area × source (Government / non-Govt)	1997-98 → present	~7k rows, AUD millions
YOUTH_JUSTICE_DETENTION	Avg nightly youth-detention pop × quarter × state × sex × Indigenous × legal status	2008 → present	~42k rows
PUBLIC_HOSPITALS	Directory of every public hospital × state × peer group × remoteness × LHN	2016-17 reference	~700 hospitals

Adding a new dataset is a single YAML drop into src/aihw_mcp/data/curated/ — see CONTRIBUTING.md.

---

Example queries (paste into Claude)

Public-health research: "For GRIM_DEATHS, give me the deaths and age-standardised rate for 'Diabetes' for Persons every year from 1980 to the latest, so I can chart the trajectory."

Health-tech / regional analysis: "Using MORT_GEOGRAPHY, list the 10 SA3 regions with the highest age-standardised mortality rate for Persons in the most recent year."

Oncology: "From CANCER_INCIDENCE_MORTALITY, give me Breast cancer incidence in Females across the 50–54 age band for every available year, plus the same age band's mortality."

Health-policy: "From HEALTH_EXPENDITURE, what was the real spend on 'Public hospitals' in NSW in 2022-23, broken down by broad source (Government vs Non-government)?"

Criminal-justice tech: "Using YOUTH_JUSTICE_DETENTION, compare the average nightly youth-detention population in NSW vs VIC in 'Jun qtr 2017', for both Indigenous and Total."

Hospital-tech / market intel: "From PUBLIC_HOSPITALS, list every 'Principal referral' hospital with their state and Local Hospital Network. Then count how many there are per state."

Each prompt resolves to one or two get_data / top_n calls. The response includes the source URL so the agent can cite it back.

---

Architecture

Same shape as the sister packages — client → cache → parsing → shaping → server:

• client.py wraps httpx with a SQLite-backed disk cache (per-resource TTL).
• parsing.py reads CSV (via pandas) and XLSX (via openpyxl/pandas). Header rows + sheet names live in the curated YAML so future format quirks are a YAML edit, not a code change.
• curated.py loads dataset specs from data/curated/*.yaml — each one declares its dimensions, measures, dimension value enums, source/download URLs, format, and parse layout.
• shaping.py transforms the parsed DataFrame into DataResponse (records / series / csv).
• server.py is the FastMCP entrypoint — six tools, full input validation with helpful "Try X" hints on error.

Cache lives under ~/.aihw-mcp/cache.db. Most AIHW datasets refresh once a year; the TTLs are tuned for that cadence.

---

Attribution

Data sourced from the Australian Institute of Health and Welfare (AIHW) via data.gov.au. Licensed under Creative Commons Attribution 3.0 Australia (CC BY 3.0 AU). The MCP server is MIT-licensed; the data carries the upstream CC-BY 3.0 AU licence, which is echoed in every response's attribution field.

---

Sister packages

• abs-mcp — ABS census and economic statistics (unemployment, CPI, GDP, population, building approvals)
• rba-mcp — RBA statistical tables (cash rate, FX rates, mortgage rates, money market)
• ato-mcp — Australian Taxation Office statistics (income by postcode, corporate tax transparency, ACNC charity register)
• aihw-mcp — this one. National mortality, regional health, cancer, expenditure, youth justice, hospitals.
• au-weather-mcp — Australian weather via Open-Meteo + BOM. 21 curated locations + postcode/place-name lookup, current observations, 16-day forecasts, 80yr historical archive.

All five are designed to compose: an agent can ask for "unemployment + cash rate + median income + mortality + climate" for postcode 2000 and one shot fans out across five MCPs.

---

Roadmap (next iterations)

• v0.2: MORT_TABLE_2 (deaths by leading cause × region); intercountry adoptions; AIHW mental-health data tables (XLSX path)
• v0.3: hosted version with x402 per-call paywall; programmatic SEO pages
• v0.4: listing on MCPay + Apify; paid tier for high-volume agent users

CHANGELOG tracks every release.

---

Development

git clone https://github.com/Bigred97/aihw-mcp.git
cd aihw-mcp
uv venv
uv pip install -e ".[dev]"
pytest                  # unit tests, no network
pytest -m live          # integration tests against data.gov.au

Issues, ideas, and contributions welcome: github.com/Bigred97/aihw-mcp/issues.