MCP server that connects Claude with Google Sheets via Apps Script
Server has 9 tools with basic descriptions in Spanish, but critical gaps in schema completeness, parameter documentation, and error handling. All tools lack structured output schemas. Parameter descriptions exist but are minimal (avg ~40 chars, well below 72-char baseline). No input validation guidance, no error recovery hints, and responses are unstructured strings rather than JSON objects. Tool names follow verb_noun pattern (good), but descriptions lack WHEN/WHY context. No tool annotations (readOnlyHint/destructiveHint). Composition is reasonable (single responsibility per tool), but output design is poor for agent chaining.
Limpia el contenido de un rango de celdas de Google Sheets
Crea una nueva hoja en el spreadsheet
Obtiene información general del spreadsheet
Lista todas las hojas disponibles en el spreadsheet
Lee el valor de una celda específica de Google Sheets
Lee un rango de celdas de Google Sheets
Prueba la conexión con Google Apps Script
No output schemas documented. Tools return unstructured strings (e.g., 'Lee el valor de una celda...') instead of JSON objects with typed fields. LLMs cannot parse results reliably or chain tools without manual extraction.
Descriptions lack WHEN/WHY context and are in Spanish. 'Lee el valor de una celda específica' (50 chars) is minimal. Should explain: when to use vs read_range, what format is returned, any prerequisites. English descriptions required for LLM clarity.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 51 | 2026-07-28+ | v2 |
Escribe un valor en una celda específica de Google Sheets
Escribe datos en un rango de celdas de Google Sheets
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Agents cannot distinguish safe reads from destructive writes without parsing descriptions. write_cell, write_range, clear_range, create_sheet should be marked destructiveHint=true.
Error responses are unstructured strings with emoji prefixes ('❌ Error', '⏱️ Error'). No error classification (retryable vs user-fixable vs fatal). No recovery guidance. 'APPS_SCRIPT_URL no está configurada' tells agent nothing about next steps.
Parameter descriptions are minimal (avg ~40 chars vs 72-char baseline). 'Nombre de la hoja (default: "demo")' lacks format constraints. No guidance on valid sheet names, cell notation rules, or range format. LLMs will guess and fail.