OpenLex MCP demonstrates solid definition quality with consistent naming patterns, comprehensive parameter schemas, and clear descriptions in German. All 8 tools follow verb_noun naming (search_*, get_*, list_*, refresh_*). Input schemas are well-structured with type constraints, minLength/maxLength bounds, and defaults. Descriptions are present and contextual (avg ~120 chars). However, output schemas are not documented, LLMs cannot predict response structure. Error handling guidance is absent; tools do not indicate what to do on failure. No tool annotations (readOnlyHint/destructiveHint) despite clear risk profiles (7 READ_ONLY, 1 WRITE). Parameter descriptions could be more prescriptive about valid formats and dependencies.

• Document output schemas for all tools. Define the structure of law objects (fields: identifier, title, status, content, etc.), search results (hits, total_count, next_cursor), and cache status (last_updated, law_count, size_bytes).
• Add tool annotations: mark all READ_ONLY tools with readOnlyHint=true; mark refresh_cache with destructiveHint=true (overwrites cache); mark search/get tools with idempotentHint=true.
• Add error handling guidance to tool descriptions. E.g., search_laws: 'Returns empty array if no matches. Try broader query or check sr_prefix filter.' get_law: 'Returns 404 if identifier not found; try list_laws() to discover valid identifiers.'
• Expand get_cache_status and refresh_cache descriptions to explain purpose: 'get_cache_status: Check when laws were last synced from HuggingFace and how many are cached locally.' 'refresh_cache: Download latest law versions from HuggingFace; call if you suspect stale data.'
• Clarify search_education_law scope in description: 'Searches only education-related laws (Volksschulgesetz, Lehrpersonalgesetz, etc.). Use search_laws for broader queries across all Zurich legislation.'

Inferred effective spec: 2025-06-18+.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry