Server has well-structured Pydantic schemas with proper type definitions and reasonable descriptions. All three tools are explicitly registered with input schemas. However, descriptions lack LLM-optimization guidance (e.g., no 'when to use' context), parameter descriptions are minimal, and output schemas are not documented. Error handling is structured but lacks recovery guidance. No tool annotations (readOnlyHint, etc.) despite all tools being read-only. Naming is clear and verb-based.

• Add output schema documentation to all three tools. Document the structure of search results (e.g., {artifacts: [{id, period, provenance, transliteration_snippet}], total_count, next_page}), metadata responses, and export formats.
• Enhance tool descriptions with 'when to use' guidance. E.g., 'Use search_artifacts to find artifacts by keyword; use get_artifact_metadata when you already have artifact IDs and need complete details.'
• Add readOnlyHint: true to all three tools in the types.Tool registration to signal safety to agents.
• Improve error responses with actionable recovery guidance. Modify _error_response() to include a 'suggested_action' field that tells the LLM whether to retry, reformulate, or escalate.
• Expand parameter descriptions with format and constraint details. E.g., 'page_size: Results per page (1-50, default 10). Larger pages reduce API calls but increase response size.' and 'fields: Comma-separated list of fields to include. Options: id (artifact ID), period (historical period), provenance (location), transliteration (cuneiform text), genre, museum_collection, bibliography.'