This MCP server exposes 15 tools for prediction market analytics. Tool naming follows verb_noun convention (status, markets, opportunities, consensus) which is good. However, critical gaps undermine quality: (1) Most tools lack input parameter descriptions, parameters like 'sport', 'limit', 'offset' are named but lack guidance on valid values, ranges, or formats. (2) Output schemas are not documented in the MCP tool definitions; TypeScript types exist in index.d.ts but are not visible to the MCP protocol layer. (3) Error handling is minimal, no recovery guidance, no categorization of retryable vs fatal errors. (4) Tool descriptions are present but generic (avg ~60 chars); they state WHAT but not WHEN to use or dependencies. (5) No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk profiles. The 'subscribe' tool is WRITE but unmarked. Baseline: 90% of A+ tools have documented return types and 100% have param descriptions, this server has neither at the MCP layer.

• Add output schema documentation to each tool definition. Include field names, types, and descriptions. Example: 'Returns {data: Market[], meta: {limit, offset, total}}'.
• Expand parameter descriptions with constraints and examples. For 'sport', specify valid values (NBA, MLS, etc.) or link to a discovery tool. For 'limit', state range (1-200, default 50).
• Add tool annotations: mark all READ_ONLY tools with readOnlyHint=true; mark 'subscribe' with destructiveHint=true (creates billing obligation).
• Implement error handling with recovery guidance. Example: 'Invalid sport. Call sports() with no args to list available sports.' Categorize errors as retryable (429, 5xx) vs user-fixable (400, 404).
• Enhance tool descriptions to 100-150 chars. Include WHEN to use: 'Use consensus() to find cross-platform agreement; use market() for single-event details.' Add dependency hints: 'Requires eventId from markets() or consensus().'.