Server has well-structured tool definitions with complete input schemas and descriptions for all 7 tools. Tool names follow verb_noun convention (criar_nota, ler_nota, etc.). Output schemas are explicitly defined via Pydantic models (NotaCompleta, ListaDeNotas, etc.). However, descriptions lack depth on WHEN to use each tool vs alternatives, and error handling guidance is minimal. Parameter descriptions are present but generic. No tool annotations (readOnlyHint/destructiveHint) despite clear risk levels. Composition is sound, tools are single-responsibility and chainable.

• Add tool annotations via FastMCP decorator metadata: @app.tool(..., readOnlyHint=True) for ler_nota, listar_notas, buscar_notas, estatisticas_base; destructiveHint=True for apagar_nota; idempotentHint=True for criar_nota (if slug is provided, it is idempotent).
• Enhance error handling: catch FileNotFoundError, ValueError, etc. in each tool and return structured error responses with actionable guidance. E.g. 'Nota não encontrada. Tente listar_notas() ou buscar_notas() para descobrir slugs válidos.'
• Add pagination to listar_notas: accept limit (1 - 100, default 20) and offset (default 0) parameters. Return total count and a next_offset field for chaining.
• Expand tool descriptions with WHEN-to-use hints. E.g. buscar_notas: 'Use this to find notes by keyword when you don't know the exact slug. For filtering by tag, use listar_notas(tag=...) instead.'
• Rename estatisticas_base to get_database_statistics or summarize_notes for clarity. Current name is too generic and does not signal the return type.