How to query Our World in Data indicators via @pipeworx/mcp-owid

Supplementary verification — @pipeworx/mcp-owid v0.1.0

Setup (library-style, NOT stdio)

npm install --prefix /tmp/pipeworx-owid @pipeworx/mcp-owid
cp /tmp/pipeworx-owid/node_modules/@pipeworx/mcp-owid/src/index.ts /tmp/pipeworx-owid/owid.ts
# Run with: node --experimental-strip-types --no-warnings
# Import: import('./owid.ts').then(mod => { const {tools, callTool} = mod.default || mod; ... })

15 calls across all 3 tools

Key NEW gotchas from this run

• ⚠️ CRITICAL: `country` requires FULL entity name — 'United States' works, 'USA' returns 0 rows silently (no error, just empty). ISO codes NOT supported despite description claiming so. Always use OWID's entity name (match against a sample fetch without country filter)
• ⚠️ Default 5000-row cap — without country filter, large datasets return first 5000 rows alphabetically (may not reach all countries). Always filter by country for targeted queries
• Temperature anomaly 2026 data already available — World +1.40°C above 1861-1890 baseline (data is near-real-time)
• Denmark 91% renewable electricity in 2025 — data validates against known wind power statistics
• 40 curated indicators across 8 categories: energy (5), climate (6), health (5), demographics (4), poverty (5), education (5), technology (5), food (5)
• Column names are machine-generated from OWID's internal schema (e.g., life_expectancy_0, emissions_total_per_capita, ny_gdp_pcap_pp_kd) — not human-friendly, check metadata first
• Some indicators include `owid_region` column (GDP does, life expectancy doesn't) — not consistent
• Invalid category returns empty list (no error), invalid slug returns 404 (clear error)

Recipe: Query Our World in Data via @pipeworx/mcp-owid

Surface: @pipeworx/mcp-owid — remote gateway at https://gateway.pipeworx.io/owid/mcp (streamable-http, JSON-RPC 2.0, SSE responses). No API key. 100 req/day anonymous cap shared across all Pipeworx packs per IP.

Package note: TypeScript source only (src/index.ts, no dist/). Cannot run locally via npx. Use the remote gateway exclusively.

Tools (3)

Catalog (40 indicators, 10 categories)

Categories: energy (5), climate (5), health (6), demographics (3), economy (4), food (4), education (3), environment (4), tech (3), politics (3).

Key slugs: life-expectancy, gdp-per-capita-maddison, co-emissions-per-capita, population, share-electricity-renewables, primary-energy-consumption-per-capita.

Calling pattern

POST https://gateway.pipeworx.io/owid/mcp
Content-Type: application/json
Accept: application/json, text/event-stream

{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"fetch_indicator","arguments":{"slug":"life-expectancy","entity":"Turkey","since_year":2000}}}

Response is SSE: parse lines starting with data: , extract result.content[0].text (JSON string).

Data shape

fetch_indicator returns tidy long-format rows: {entity, year, value}. Country names work as-is ("Turkey", "United States", "World"). since_year/until_year filter works. Invalid slugs return graceful JSON error with retry_hint.

get_indicator_metadata returns: description, unit, source, lastupdated, nextupdate.

Verified traces (10 calls, 100% success)

1. list_popular_indicators({category:"energy"}) → 5 indicators
2. list_popular_indicators({category:"health"}) → 6 indicators
3. list_popular_indicators({}) → 40 indicators across 10 categories
4. fetch_indicator({slug:"life-expectancy",entity:"Turkey",since_year:2000}) → 24 rows, 71.9→77.2 years (2000-2023), COVID dip visible at 2020
5. fetch_indicator({slug:"gdp-per-capita-maddison"}) → 20 rows of global GDP data
6. fetch_indicator({slug:"co-emissions-per-capita",entity:"United States",since_year:1990,until_year:2023}) → 34 rows, 20.25→~14.8 tonnes (27% decline)
7. fetch_indicator({slug:"population",entity:"World",since_year:2020}) → 4 rows, 7.89B→8.09B (2020-2023)
8. fetch_indicator({slug:"nonexistent-indicator-xyz"}) → graceful 404 error with retry_hint
9. get_indicator_metadata({slug:"life-expectancy"}) → unit=years, sources=Riley+HMD+UN WPP
10. get_indicator_metadata({slug:"co-emissions-per-capita"}) → unit=tonnes, source=Global Carbon Budget

Gotchas

• No local execution: package has no compiled dist, only TS source. Must use remote gateway.
• 100 req/day cap: shared across ALL Pipeworx packs per IP. Budget calls across owid, dbnomics, etc.
• SSE response format: gateway returns event: message\ndata: {jsonrpc response} — parse the data: line.
• Invalid slugs: returns structured error, not MCP error code. Check for error field in parsed content.