MCP server adaptation for Quran Search Engine - provides tools to search the Quran with support for lemma, root, and fuzzy matching
The Quran Search Engine MCP has one tool ('search') with a reasonably complete JSON Schema input definition and a basic description. The schema includes proper type definitions and parameter descriptions for most fields, meeting baseline requirements. However, the description is terse (21 chars), parameter descriptions are minimal and formulaic, and there is no documented output schema. The tool lacks error handling guidance, recovery patterns, and does not follow LLM-optimized description practices. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) are present despite the tool being read-only. The single-tool design is appropriate for a focused domain (Quran search), but the definition quality falls short of production-grade standards.
Search the Quran
Tool description is too brief (21 characters). Does not meet 50 - 200 character baseline for LLM optimization. Missing context on when to use the tool, what data it returns, and any prerequisites.
No output schema documented. The tool accepts query and options, but the LLM cannot know what fields to expect in the response (e.g., does it return verse_number, text, surah_name, morphology?). This forces the LLM to guess the shape of results.
No error handling or recovery guidance. If the search fails (invalid query, no results, malformed morphology request), the tool provides no hint to the LLM on what to do next (retry? simplify query? check parameter format?).
Tool lacks read-only annotation. Despite the tool being a read-only search operation, it does not declare readOnlyHint=true. This prevents the MCP client from determining whether the tool is safe for automated agents.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 52 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 11 | - | v1 |
Parameter descriptions are minimal and formulaic ('The search query string', 'Optional search options', 'Maximum number of results to return'). They do not provide actionable guidance on expected format, valid ranges, or constraints. For example, 'query' should specify minimum/maximum length and whether special characters (Arabic diacritics, hamza) are supported.
matchType enum ('exact', 'fuzzy', 'prefix') lacks descriptions explaining the difference and when to use each. An LLM must infer intent from the enum name alone.