MCP server for INEGI's open data — search Mexico's national business directory (DENUE, 5M+ establishments) and query official socioeconomic indicators (population, inflation, and more) from Claude and other AI agents.
Server defines 5 tools with consistent naming (verb_noun pattern: denue_buscar_*, denue_detalle, inegi_indicador). All tools have descriptions (194 - 280 chars, within baseline 34 - 392 range). Input schemas use Zod with type constraints and descriptions for all parameters. However, output schemas are not documented, responses are formatted as plain text summaries via resumen() helper, not structured objects. Parameter descriptions are adequate but lack actionable format guidance (e.g., 'estado' accepts state names OR 2-digit codes, but the mapping is opaque to the LLM). No error handling guidance, failures return generic text. Tool composition is sound (each does one thing), but output lacks chaining IDs and structured fields needed for multi-step agent workflows.
Busca establecimientos economicos de todo Mexico en el Directorio Estadistico Nacional de Unidades Economicas (DENUE) de INEGI, cerca de una coordenada dada. Util para encontrar negocios, competidores o proveedores en una zona especifica.
Busca establecimientos economicos por palabra clave dentro de un estado especifico de Mexico (o a nivel nacional). Util cuando no se tiene una coordenada exacta, solo una region.
Busca establecimientos economicos por su nombre comercial o razon social en todo Mexico o en un estado especifico.
Obtiene la ficha completa de un establecimiento economico dado su ID de DENUE (obtenido de otra busqueda).
Consulta indicadores socioeconomicos oficiales de INEGI (Banco de Indicadores): poblacion, inflacion, confianza del consumidor, actividad industrial, y mas, a nivel nacional o por estado. Indicadores conocidos por nombre: poblacion, inflacion, inpc, confianza del consumidor, actividad industrial. Tambien acepta directamente un codigo numerico de indicador de INEGI para consultar cualquiera de su catalogo.
Output schemas not documented. All tools return plain-text summaries via resumen() helper. LLMs cannot parse structured fields, extract IDs for chaining, or validate response shape. Responses should be JSON objects with typed fields (id, nombre, actividad, direccion, etc.).
Parameter 'estado' accepts both state names ('Jalisco') and 2-digit codes ('14'), but the mapping logic (resolveEstadoCode) is hidden in implementation. Description should explicitly list valid formats and examples, or provide an enum of state names.
No error handling guidance. If a search returns zero results, the tool returns 'No se encontraron establecimientos...' as plain text. If the API fails, the LLM receives an unstructured error. Errors should categorize as retryable/user-fixable and suggest recovery steps.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 69 | 2026-07-28+ | v2 |
No pagination support in responses. denue_buscar_* tools accept 'inicio' and 'fin' parameters (max 100 results per call), but responses do not include total_count, next_cursor, or has_more flags. LLMs cannot determine if more results exist or how to fetch them.
Tool descriptions lack dependency hints. E.g., denue_detalle requires an 'id' from a prior search, but the description does not say 'Call denue_buscar_* first to obtain an ID.' This forces LLMs to infer the workflow.