Universal analytics MCP server supporting Google Analytics 4 (via GA4 Data API) and simple CSV analytics with demo mode
Four tools with basic verb-noun naming (get_*) and present descriptions, but significant gaps in parameter documentation and output schema clarity. All tools accept 'days' parameter with type and description, but lack enum constraints, range limits, and output schema documentation. Descriptions are adequate (50-80 chars) but generic, they state WHAT without WHEN or WHY. No error handling guidance, no pagination despite returning lists, no field-level documentation for complex outputs. Tools return JSON strings rather than structured objects, forcing LLM parsing. No tool annotations (readOnlyHint, idempotentHint). Composition is reasonable (each tool has one responsibility), but chaining is weak, no IDs returned to enable downstream calls.
Get a comprehensive analytics summary combining all metrics.
Get top pages by view count.
Get traffic sources and their conversion rates.
Get user demographics: countries, devices, age groups.
Output schemas not documented. Tools return JSON strings (via json.dumps) with no declared structure. LLMs cannot plan downstream calls or extract fields reliably. E.g., get_page_views returns {period, pages, note} but this is inferred from code, not declared in tool metadata.
Parameter 'days' lacks range constraints. No min/max documented. LLMs could pass 0, negative, or absurdly large values (e.g., 999999 days). Rubric baseline: numeric params should specify min - max (e.g., days 1 - 365).
No pagination support. get_page_views accepts 'limit' but returns all pages in a single response with no next_cursor, total_count, or offset. Large result sets will blow context windows. Rubric baseline: tools returning lists should support pagination.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 57 | 2026-07-28+ | v2 |
No error handling guidance. If GA4 API fails, get_ga4_data returns {error, fallback} but tools do not document what errors are possible, whether they are retryable, or what the LLM should do next. Rubric: error responses must tell the LLM what to do next.
No tool annotations. Tools lack readOnlyHint, idempotentHint, or destructiveHint. All four are read-only and idempotent, but this is not declared in metadata. Clients cannot optimize caching or retry logic without explicit hints.