Server exposes 7 code-intelligence tools with reasonable naming (verb_noun pattern: semantic_search, code_search, symbol_lookup, file_outline, find_callers, index_status, open_workspace). All tools have descriptions and input schemas visible in main.rs. However, descriptions are terse (10-50 chars), parameter descriptions lack detail on constraints/formats, and output schemas are not documented. No error handling guidance, no tool annotations (readOnlyHint/destructiveHint), and no pagination support despite search tools returning potentially large result sets. Schemas present but minimal, parameters lack examples, ranges, or validation rules.

• Expand tool descriptions to 100-200 characters. Include WHEN to use (e.g., 'Use semantic_search for natural-language queries about functionality; use code_search for exact identifiers or signatures'), WHAT it returns (e.g., 'Returns matching symbols with file paths and line numbers'), and dependencies (e.g., 'Requires workspace to be indexed via open_workspace first').
• Add explicit parameter constraints: semantic_search.limit should be '1-100 (default 15)', code_search.limit '1-100 (default 20)'. Document workspace parameter as 'Absolute path; if omitted, uses first indexed workspace'.
• Document output schemas for all tools. Example for semantic_search: 'Returns array of {symbol_name: string, file_path: string, line_number: integer, confidence: number 0-1, context: string}'.
• Add pagination to search tools: add offset/limit or cursor-based pagination. Cap default results at 20-30 and document in description: 'Returns up to 20 results; use offset parameter for additional pages'.
• Add tool annotations via MCP schema: mark semantic_search, code_search, symbol_lookup, file_outline, find_callers, index_status with readOnlyHint=true; mark open_workspace with destructiveHint=true (modifies workspace state).

Inferred effective spec: <=2025-11-25.

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