Server has 6 tools with explicit schemas and descriptions. All tools are READ_ONLY except submit_sitemap (WRITE). Schemas are properly defined using Zod and converted to JSON Schema via zodToJsonSchema. However, descriptions are generic and lack actionable context for LLM selection. Parameter descriptions exist but are minimal (avg ~80 chars). No output schema documentation. Error handling is basic (Zod validation + generic catch). Tool names follow verb_noun convention (list_, search_, get_, submit_) which is good. Missing: output structure docs, error recovery guidance, parameter constraints (enums where applicable), and LLM-optimized descriptions that explain WHEN to use each tool.

• Expand tool descriptions to 100-200 chars explaining WHEN to use each tool and what it returns. E.g., 'search_analytics: Retrieve search performance metrics (impressions, clicks, CTR, position) for a site. Call list_sites() first to get valid siteUrl format. Returns rows grouped by specified dimensions (query, page, country, device, searchAppearance).'
• Document output schemas for each tool. Add a comment block or separate schema file showing the structure of response.data for each tool. E.g., search_analytics returns {rows: [{query, clicks, impressions, ctr, position}], responseAggregationType}.
• Add enum descriptions to search_analytics parameters. Explain what each 'type' value filters (web=web search results, image=Google Images, video=Google Videos, news=Google News). Similarly, explain aggregationType options.
• Implement recovery-guided error messages. When Zod validation fails, catch the error and return a helpful message like 'Invalid siteUrl format. Valid formats: sc-domain:example.com (domain property) or http://www.example.com/ (site prefix). Call list_sites() to see your registered properties.'
• Add pagination support to search_analytics. Document the rowLimit parameter (max 25000 per Google API docs), clarify whether results are truncated, and provide guidance on fetching additional rows if needed.