Bible MCP demonstrates solid tool design with clear naming conventions, well-structured schemas using Zod, and comprehensive descriptions. All 7 tools follow verb_noun naming patterns (get_verse, search_bible, list_books, etc.). Descriptions are detailed and provide context for LLM selection. Input schemas are properly typed with Zod validators. However, output schemas are not explicitly documented in the visible code, responses are constructed ad-hoc as text concatenation rather than returning structured objects. Error handling exists but lacks actionable guidance for LLM recovery. The read_bible tool stands out as an outlier, returning HTML rather than structured content, which may confuse LLM parsing. Security considerations are adequate (read-only operations, no secrets exposed), but parameter descriptions could be more concise to avoid token waste.

• Document explicit output schemas for all tools. For get_verse, return: {reference: string, translation: {id, name}, verses: [{book, chapter, verse, text}]}. Use the interface definitions (VerseResponse, ChapterResponse, etc.) already in code but surface them via MCP output documentation.
• Replace read_bible's HTML response with structured JSON containing verse objects, navigation metadata, and translation info. Remove or delegate the HTML viewer to a separate resource if UI rendering is needed.
• Enhance error responses with recovery guidance. E.g., if reference is invalid, return: 'Invalid reference: got "John 99:99". Try list_books() to see available books, or use format like "John 3:16" (Book Chapter:Verse).'
• Use Zod enum constraints for translation IDs instead of repeating them in descriptions. Define const TRANSLATION_IDS = z.enum(['web', 'kjv', 'wlc']) and use z.string().refine(val => TRANSLATION_IDS.safeParse(val).success) with a brief description.
• Add tool annotations to all tools: registerAppTool(..., {readOnlyHint: true}). This clarifies safety contracts to callers.
• For list_translations(), add optional parameters: language (string) and license (enum: 'public-domain', 'cc-by', etc.) to enable discovery filtering.
• Document search_bible filter interaction: 'Filters combine with AND logic: if both book and testament are provided, results are limited to verses in that specific book within that testament.'
• Consider pagination for search_bible results: add 'next_cursor' or 'page_token' to response to support large result sets without truncating.