The server has 6 tools with basic descriptions but critical gaps in schema documentation and parameter specification. Tool descriptions are present but generic and lack actionable guidance for LLMs. No visible input parameter schemas or output schema documentation in the provided source code. Tool names use verb-first patterns (search-*, code-*, help-*) which is good, but there's no evidence of detailed JSON Schema definitions, parameter type annotations, or error recovery guidance. The code references 'main_mcp_server.py' but the actual tool registration and schema definitions are not visible in the provided source, making it impossible to verify schema quality or validate that parameters have proper types and descriptions.

• Add explicit JSON Schema definitions for all tool inputs. Each tool must declare parameters with 'type', 'description', and 'enum' (for constrained values). Example: 'query' parameter for search tools should be type 'string' with min/max length constraints.
• Document output schemas. Specify what fields are returned for each tool, their types, and whether they are paginated. For search tools, include 'total_count', 'results' (array), and per-result fields like 'id', 'score', 'content'.
• Expand tool descriptions to 80-150 characters explaining WHAT the tool does, WHEN to use it vs alternatives, and WHAT it returns. Example: 'Perform vector similarity search over indexed code. Use when the query is conceptual ("error handling patterns"). Returns ranked results with relevance scores. For keyword matching, use search-keyword instead.'
• Move inline syntax documentation (e.g., keyword query syntax in search-hybrid) into a structured 'syntax' parameter description or a dedicated help tool result. The search-hybrid description should focus on the high-level use case, not implementation details.
• Add error handling examples. Document what the LLM should expect if: (1) the embedding service is unavailable, (2) no results match, (3) the index is empty. Provide recovery steps like 'If embedding fails, try search-keyword instead' or 'If no results, try a broader query'.
• Clarify differences between search tools in their descriptions. Currently, search-vector vs search-keyword vs search-hybrid are unclear. Example: search-vector='Pure semantic similarity (ignores keywords)', search-keyword='Exact field matching (ignores meaning)', search-hybrid='Combines both for best recall'.

Inferred effective spec: 2025-06-18+.

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