MCP server for FAO SDMX statistical data access, providing tools to query dataflows, structures, constraints, and data from FAO statistical databases
Server has 8 well-structured tools with explicit schemas and descriptions. All tools are READ_ONLY and properly registered via registerTool(). Naming follows verb_noun convention (sdmx_*). Descriptions are present and substantive (100-250 chars typical). Input schemas use Zod with type definitions and defaults. However, parameter descriptions are sparse or missing entirely, most params lack inline documentation explaining their purpose, constraints, or expected format. Output schemas are defined but not visible in the excerpt. Error handling is minimal; no recovery guidance or actionable error messages. No tool annotations (readOnlyHint, etc.) despite all being read-only.
Fetch /availableconstraint for a dataflow and return the raw JSON plus a compact allowed-values summary when possible. Auto-selects datasource when datasourceId is omitted.
List known FAO SDMX datasources and base URLs.
List dataflows via GET /dataflow/{agency}/ (same REST endpoint used by Domain Studio BFF). Supports optional search filter.
Resolve a human-friendly selection (e.g. {SPECIES: 'tuna', REF_AREA: 'Italy'}) to SDMX codes and fetch matching data rows. Supports fuzzy code matching and vernacular overlays.
Specialized fishery query: resolve species and area codes, then fetch and sum capture + aquaculture production by year.
Execute a full SDMX data query with explicit dimension selection and optional time filtering. Returns normalized rows plus optional raw payload and structure.
Parameter descriptions missing or minimal. Params like 'agency', 'datasourceId', 'baseUrl', 'selection', 'locale', 'maxRows' lack inline documentation explaining their purpose, valid ranges, or format constraints. Rubric baseline: 100% of A+ tools have param descriptions.
No tool annotations despite all tools being read-only. Missing readOnlyHint annotations would help LLMs understand safety and retry semantics. Current spec (2026-07-28) rewards tool annotations.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | 2026-07-28+ | v2 |
Search codelist labels within a dataflow dimension (e.g. SPECIES=tuna, REF_AREA=Italy). Uses multilingual labels, scored ranking, and fishery vernacular overlay. Auto-selects datasource when datasourceId is omitted.
Fetch SDMX structure via /dataflow/{agency}/{dataflow}/{version}?references=all&detail=full and parse dimensions/codelists. Use lite=true for summary only (no raw payload). Auto-selects datasource when datasourceId is omitted.
Error handling is implicit. No explicit error recovery guidance, categorization (retryable vs fatal), or actionable error messages. Tools that call external APIs (fetchDataflowCatalog, fetchStructure, etc.) may fail but responses do not guide LLM recovery.
Output schemas defined in separate file (outputSchemas.js) but not visible in excerpt. Cannot verify if outputs include chaining IDs, pagination metadata, or follow response-shaper pattern. Rubric baseline: 100% of A+ tools have documented return types.
Parameter 'selection' in sdmx_availableconstraint and sdmx_query accepts additionalProperties but lacks description of expected structure, valid keys, or format. LLMs cannot infer how to construct this object.