Server has 6 tools with complete input schemas and descriptions. Naming follows verb_noun pattern (validate_*, generate_*, parse_*, check_*). Descriptions are present but vary in quality (34 - 194 chars). All parameters have type and description. However, output schemas are not documented in the source code, parameter constraints lack enums/ranges, and error handling guidance is minimal. No tool annotations for destructive operations despite two WRITE-risk tools (generate_xrechnung, generate_zugferd). Composition is sound, each tool has one responsibility. Missing: pagination guidance, output field documentation, recovery hints in error messages.

• Document output schemas for all 6 tools. Example: 'Returns {errors: [{code, message}], warnings: [{code, message}], profile: string}'. This enables agents to extract and chain results.
• Add enum constraints to country_code parameters (ISO 3166-1 alpha-2: DE, AT, CH, etc.) and replace 'JSON array' description for items with a formal schema or example structure.
• Declare tool annotations: add destructiveHint=true to generate_xrechnung and generate_zugferd so agents know these modify state.
• Split parse_einvoice and check_compliance into separate tools for XML and PDF inputs (parse_xrechnung, parse_zugferd, check_xrechnung_compliance, check_zugferd_compliance) to eliminate type ambiguity.
• Add error recovery guidance to validation tool descriptions: 'Returns validation errors. If errors exist, review against KoSIT rules at https://www.kosit.bund.de/. Retry after fixing XML structure.'

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

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