Rootdata MCP has 17 tools with complete JSON Schema definitions and descriptions present for all tools. However, quality is inconsistent. Tool names lack clear verb-noun structure (e.g., 'analyzeComprehensive' vs 'analyze_project', 'investigateEntity' vs 'investigate_entity'). Descriptions are present but generic, most are 50-120 chars, below the 194-char baseline for production tools. Parameter descriptions exist but lack actionable constraints: no ranges for numeric params (page, pageSize, days), no format guidance for timestamps, no enum validation hints in descriptions. Output schemas are not documented, LLMs cannot predict response structure. Error handling is minimal; no recovery guidance or actionable error messages visible in code. Type guards exist but validation is loose. Overall, definitions are structurally sound but lack the depth and precision needed for reliable LLM reasoning.

• Rename tools to snake_case verb_noun format: 'analyzeComprehensive' → 'analyze_project', 'investigateEntity' → 'investigate_entity', 'trackTrends' → 'list_trends'. This matches the 90% baseline and helps LLMs parse intent from names alone.
• Document output schemas in tool descriptions. Example: 'Returns {project_id, name, funding_usd, investor_ids[], ecosystem_id}. Use investor_ids to call getInvestors() for details.' This enables tool chaining and prevents discovery detours.
• Add numeric constraints to parameter descriptions. Change 'Number of results per page' to 'Number of results per page (1 - 100, default 10)'. Change 'Number of days to look back' to 'Number of days (1 - 365, default 30)'. This prevents LLMs from passing invalid values.
• Expand tool descriptions to 150 - 250 chars. Add WHEN to use and dependencies. Example: 'Search for projects by name or keyword. Returns top 20 results with IDs. Use getProject(projectId) for full details. Call this first if you only have a project name, not an ID.'
• Implement structured error responses with recovery guidance. Catch API errors and return: {error: 'Project not found', suggestion: 'Try searchProjects() with a partial name', retryable: false}. This guides LLM recovery without extra calls.