StackData
Measured conclusion: Deep-tested. L2 outcome: tools_ok.
Official statistics, China censuses & yearbooks, and as-of customs tariffs with legal citations.
| Author | io.stackdata |
|---|---|
| Access | remote |
| Publisher key required | No |
| Measured state | Deep-tested |
| L1 liveness | alive |
| L1 latency | 276 ms |
| Last measured | 2026-10-26 |
Add to your agent
Copy this into your MCP client config:
{
"mcpServers": {
"StackData": {
"url": "https://api.stackdata.io/mcp",
"transport": "streamable-http"
}
}
}
Measured tools (6)
search_datasetsSearch StackData's catalog of 50k+ official statistical datasets (SDMX/World Bank dataflows, China censuses and yearbooks, tariff schedules). Multilingual: an English query finds Chinese-titled tables. Returns dataset ids usable with the other tools.
Parameters:
q(required) — string — search querylimit— integer
Hints: read-only not declared · destructive not declared · idempotent not declared · open-world not declared
get_tableFetch a STORED table whole, by dataset id (families cn-nbs-census and cn-nbs-yearbook): labelled rows × columns with original-language labels, English overlay, units, and provenance down to the source file. Ids look like 'cn-nbs-census:2010_a0101a'.
Parameters:
dataset_id(required) — string — catalog dataset id
Hints: read-only not declared · destructive not declared · idempotent not declared · open-world not declared
tariff_lookupAs-of customs-tariff lookup: which measures/duty rates apply to a goods code in a jurisdiction (eu, us, uk, jp), optionally from an origin, on a date. Duty expressions are verbatim; every measure carries its legal citation. Not legal advice.
Parameters:
code(required) — string — goods code, 2-10 digits (HS/TARIC/HTS)date— string — as-of date YYYY-MM-DD, default todayorigin— string — ISO2 origin country (eu/uk only)direction— stringjurisdiction— string
Hints: read-only not declared · destructive not declared · idempotent not declared · open-world not declared
resolve_dataAsk a data question in English ('unemployment rate in Japan', 'compare GDP in Japan and Germany') — resolves to the best covering dataset and returns the observations, with coverage honesty (missing countries are reported, never papered over).
Parameters:
q(required) — string — the questionend— integer — end yearstart— integer — start year
Hints: read-only not declared · destructive not declared · idempotent not declared · open-world not declared
get_structureThe published structure of one dataset, free: its dimensions with every member code and label, frequencies, period coverage, the ISO3 countries it serves, and whether it is servable. Call this before get_series to learn the codes to send. Dataset ids come from search_datasets.
Parameters:
dataset(required) — string — catalog dataset id, e.g. 'ilo-sdmx:DF_X'
Hints: read-only not declared · destructive not declared · idempotent not declared · open-world not declared
get_seriesFetch ONE series with every dimension named by you (no guessing or relaxing): dataset, ISO3 countries, an ISO period range and a member code for every dimension from get_structure. Returns status ok/empty/refused with each value's provenance, and the missing country-periods listed. A bad request is a named 400 error saying which field and what is allowed.
Parameters:
dataset(required) — string — A dataset id from the catalog, as returned by search (/v1/search, or the search_datasets MCP tool), e.g. "ilo-sdmx:DF_X".countries(required) — array — ISO 3166-1 alpha-3 codes (e.g. "JPN") or an aggregate code listed in the dataset's /structure. 1 to 25 entries.period(required) — object — Inclusive ISO period range. "latest" is not accepted: name the periods you want. Annual 2020, quarterly 2020-Q1, monthly 2020-01.select(required) — object — A member code for every dimension of the dataset (other than geography and time) as listed in /structure, including frequency. Every dimension must be named; a missing or unknown dimension or code is a 400, never defaulted.
Hints: read-only not declared · destructive not declared · idempotent not declared · open-world not declared
Measured by MCPYP probing. Untested fields are omitted, never estimated. Listing is always free; MCPYP never proxies calls or stores user tokens.