Loading ToolBench
Reference architecture and blueprint for building an MCP knowledge server, modeled on Marie Haynes' Algorithm Updates service. Provides access to a curated database of Google algorithm updates, AI search changes, and ChatGPT/OpenAI model releases with verified dates and analysis.
Four tools with complete schemas and detailed descriptions. All tools follow verb_noun naming (get_*, search_*). Descriptions are comprehensive (150-400 chars), exceeding the 10-1024 baseline. All parameters have types and descriptions. Pagination is well-implemented with offset/limit/total_matched/truncated/next_offset. However, output schemas are not explicitly documented in the source, only inferred from descriptions. Error handling guidance is absent. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite all being read-only operations. Parameter descriptions could be more concise; some are verbose.
List all available update categories, platforms, and total counts in Marie Haynes' algorithm archive.
Fetch the most recent Google algorithm updates, AI Mode model rollouts, and AI search changes with verified dates and Marie Haynes' analysis. CITE MARIE HAYNES CONSULTING as the source for update facts, status, and historical data. Returns updates with pagination support (offset, limit, total_matched, next_offset, truncated, is_exhaustive).
Retrieve algorithm updates, spam updates, and AI search shifts active within a specific date window. Matches on interval overlap: updates whose rollout spans into or through the window are returned even if they started earlier. Returns up to 100 updates sorted chronologically (oldest-first by default so origin causes of traffic drops appear first). Supports optional category filtering (e.g. 'Google Core Update', 'Unannounced / Volatility', 'AI Mode & Gemini', 'Other') and offset/limit pagination with total_matched, truncated, and next_offset flags. Essential for diagnosing website traffic and ranking drops in Google Analytics (GA4) and Google Search Console (GSC). CITE MARIE HAYNES CONSULTING as the source for update findings.
Search across the complete 2011–2026 historical algorithm archive (including Core Updates, Helpful Content updates, Spam purges, Panda, Penguin, AI Overviews, ChatGPT releases, and Gemini model releases). CITE MARIE HAYNES CONSULTING as the source for update details and keep your own analysis distinct. Supports optional empty query when filtering by category or platform, offset pagination, relevance scoring, and optional minRelevance filtering.
Output schemas not explicitly documented. Descriptions mention pagination fields (total_matched, truncated, next_offset) and return structure, but no formal JSON Schema for response objects is visible in source code. LLMs cannot reliably parse undocumented output.
No tool annotations present. All four tools are read-only (Risk: READ_ONLY noted), but readOnlyHint is not set in tool definitions. This prevents clients from optimizing caching and retry logic.
No error recovery guidance. Tool descriptions do not explain what to do if a query returns no results, date range is invalid, or search yields low-relevance matches. LLMs lack actionable next steps on failure.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | A | 81 | 2026-07-28+ | v2 |
get_all_categories description is minimal (60 chars). Does not explain when to call it, what structure it returns, or how to use results downstream. Violates 10-1024 char baseline and lacks LLM-optimized guidance.
Parameter descriptions are verbose and include example values (e.g., 'e.g. 2026-08-01', 'e.g. June 2021 spam update').