ainu-mcp provides 10 tools with consistent naming (verb_noun pattern: dictionary_lookup, corpus_search, morpheme_search, etc.) and reasonable descriptions (avg ~80 chars). All tools have input schemas with typed parameters and descriptions. However, output schemas are not documented in the source code, parameter descriptions lack detail about constraints/formats, and error handling guidance is absent. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications. The server targets a specialized linguistic domain with well-scoped tools, but lacks production-grade polish in schema completeness and error recovery patterns.

• Document output schemas for all tools. For each tool, add a 'Returns' section to the description or a separate schema field specifying response object structure, field names, types, and example values. E.g., dictionary_lookup returns {entry_id, word, definition, part_of_speech, examples: [{sentence, translation}]}.
• Expand parameter descriptions to 50-200 characters with explicit constraints. E.g., 'query: Ainu word or phrase (1-100 chars, Latin script or Katakana). Partial matches supported.' Include format hints and valid ranges.
• Add tool annotations via MCP SDK. Mark glossary_add and glossary_update with destructiveHint=true; mark all READ_ONLY tools with readOnlyHint=true. This enables clients to display safety warnings and enforce permission checks.
• Add error recovery hints to lookup tools. E.g., 'If the exact term is not found, try corpus_search() with a partial phrase or morpheme_search() to find related forms.' This guides multi-step agent planning.
• Clarify optional parameter defaults in descriptions. E.g., 'dictionary: Optional. If omitted, searches all available dictionaries (Ainu-Japanese, Ainu-English, etc.). Specify a dictionary name to narrow results.' Same for dialect in corpus_search.