Three tools with clear verb-noun naming (get_*) and complete input schemas. Descriptions are present and adequate (100-150 chars each), explaining WHAT and WHEN. Parameters have types and descriptions. However, output schemas are undocumented (tools return generic `str`), and error handling lacks recovery guidance. No tool annotations (readOnlyHint, etc.). Parameter constraints are minimal, max_reviews lacks bounds in get_reviews; sort enum values not formally constrained. Composition is sound: tools are single-purpose and chainable via org_id. Missing: structured output documentation, error classification, and LLM-optimized descriptions.

• Document output schemas for all tools. Define the JSON structure returned (e.g., reviews array with fields: text, author, date, rating, response). Use Pydantic models or JSON Schema to make output types explicit.
• Add formal enum constraints to sort parameter: declare as enum=['by_time', 'by_rating'] in schema. Add min/max bounds to max_reviews (e.g., 1 - 100).
• Implement tool annotations via FastMCP. Mark all three tools with readOnlyHint=true since they are read-only. This signals to clients that tools are safe to call without confirmation.
• Add error handling with recovery guidance. Document common errors (e.g., 'Invalid org_id', 'Rate limit exceeded') and suggest next steps (e.g., 'Verify org_id from Yandex Maps URL' or 'Wait 60 seconds before retrying').
• Enhance descriptions with dependency hints and workflow context. E.g., 'Use get_company_summary for a quick overview with recent reviews (default 10). For full review history or custom sorting, call get_reviews with max_reviews and sort parameters.'

Inferred effective spec: <=2025-11-25.

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