CodeKurve MCP provides 14 well-structured code analysis tools with consistent naming (verb_noun pattern: search_symbols, find_references, trace_path) and comprehensive input schemas. All tools have descriptions (avg ~120 chars) and typed parameters with descriptions. However, output schemas are not documented in the visible source, parameter descriptions lack constraint details (ranges, enums, formats), and error handling guidance is absent. Tools follow a read-only pattern (13/14) with one write operation (reindex), but no confirmation or dry-run patterns are evident. Schemas are present and valid JSON Schema, but descriptions could be more LLM-optimized per the 50-200 char baseline.

• Document output schemas for all tools. For search_symbols, specify: {results: [{id: string, name: string, kind: string, file: string, line: number}], total: number}. For find_references, specify confidence levels and reference types returned.
• Formalize min_confidence as an enum in the schema: {type: 'string', enum: ['exact', 'high', 'medium', 'low', 'unresolved']} instead of prose description.
• Add min/max bounds to numeric parameters: limit (1-1000, default 20), offset (0+), ctx_lines (0-50), depth (1-10). Include defaults in descriptions.
• Add error recovery guidance to tool descriptions. E.g., 'If symbol not found, try codekurve_search_symbols() with a partial name. If index is stale, call codekurve_doctor() to check freshness.'
• For reindex tool, add description: 'Rebuilds the code graph index. This is a blocking operation that may take minutes on large projects. Only call if codekurve_doctor() reports staleness. Requires [mcp] allow_reindex = true in config.'