Query UNICEF child health, education, and WASH statistics via @pipeworx/mcp-unicef (library-style) -- 3 tools, credential-free

@pipeworx/mcp-unicef v0.1.0 -- UNICEF Child Statistics via SDMX 2.1 REST API

Library-style module (NOT a stdio server). Install: npm install @pipeworx/mcp-unicef, then copy the TS source out of node_modules/@pipeworx/mcp-unicef/src/tools/unicef.ts and run with node --experimental-strip-types --no-warnings --input-type=module.

The package exports {tools, callTool} — call callTool(name, args) directly; no MCP stdio transport needed.

3 Tools

Key Gotchas

1. Dataflow IDs are SHORT codes, not descriptive names. CME = child mortality estimates, NUTRITION = nutrition, IMMUNISATION = immunisation. Use list_dataflows to discover them.
2. WASH is split into multiple dataflows: WASH_HEALTHCARE_FACILITY, WASH_HOUSEHOLDS, WASH_SCHOOLS, WASH_COMMUNITIES, WASH_COMMUNITY. Searching 'water' returns 0 results; search 'WASH' instead.
3. Global queries are EXTREMELY SLOW — get_data({dataflow_id:'CME'}) with no key filter takes ~90 seconds and returns 21,872 series. ALWAYS filter by REF_AREA (country ISO code) or use max_series to cap.
4. Key format is dot-separated dimension positions — e.g. for CME with 4 dimensions (REFAREA, INDICATOR, SEX, WEALTHQUINTILE), the key USA... means USA + all indicators + all sexes + all quintiles. Empty segments = wildcard.
5. DSD dimension inspection is essential — each dataflow has different dimensions in different positions. Call dataflow_structure before constructing keys.
6. `last_n` parameter returns only the N most recent observations per series — useful for getting latest data without date range guessing.
7. 404 on wrong dataflow_id — the API returns HTTP 404 with no helpful message if the ID doesn't exist. Always verify with list_dataflows first.
8. limit parameter on list_dataflows caps results — default 20, max seems unbounded but practical API returns vary. Search with specific terms for best results.
9. Nutrition indicators include stunting, wasting, overweight, underweight, breastfeeding — the NUTRITION dataflow has ~40 indicators across 200+ countries.
10. Period format is YYYY (annual) for most dataflows — e.g. start_period: '2020', end_period: '2023'.

Verified Execution Trace (~18 calls, 11 OK + 2 correct rejections + 1 wrong-ID, p50=675ms)

list_dataflows({}) -> 20 dataflows: CME, NUTRITION, IMMUNISATION, EDUCATION, HIV_AIDS, CHILD_LABOUR... (1204ms)
list_dataflows({query:'mortality'}) -> 1 hit: CME (Child Mortality Estimates) (643ms)
list_dataflows({query:'WASH'}) -> 5 hits: WASH_HEALTHCARE_FACILITY, WASH_HOUSEHOLDS, WASH_SCHOOLS, WASH_COMMUNITIES, WASH_COMMUNITY (708ms)
list_dataflows({query:'water'}) -> 0 hits (587ms)
list_dataflows({query:'education'}) -> 3 hits: EDUCATION, EDUCATION_PARITY, EDUCATION_PARTICIPATION (694ms)
dataflow_structure({dataflow_id:'CME'}) -> 4 dimensions: REF_AREA(266 codes), INDICATOR(31 codes: CME_MRY0T4/CME_MRM0/CME_MRY0/...), SEX(3: _T/M/F), WEALTH_QUINTILE(6) (1529ms)
dataflow_structure({dataflow_id:'NUTRITION'}) -> dimensions: REF_AREA, INDICATOR(~40), SEX, RESIDENCE, WEALTH_QUINTILE (1120ms)
get_data({dataflow_id:'CME', key:'USA...', last_n:3}) -> Under-5 mortality USA: 2021=6.4, 2022=6.5, 2023=6.5 per 1000; Neonatal: 2023=3.3; Infant: 2023=5.2 (2834ms)
get_data({dataflow_id:'CME', key:'TUR...', last_n:3}) -> Under-5 Turkey: 2021=9.4, 2022=8.7, 2023=8.1 per 1000 live births (1927ms)
get_data({dataflow_id:'CME', key:'NGA...', last_n:2}) -> Unde