Server has 9 tools with complete input schemas and descriptions. Naming follows verb_noun convention (top_markets, search_markets, get_market, etc.). Descriptions are present and adequate (avg ~80 chars), but lack actionable context about when to use each tool vs. alternatives. Parameters have types and defaults, but lack detailed constraints (e.g., no min/max for limit, no enum for tag values). Output schemas are not documented, LLMs cannot predict response structure. Error handling is basic (retry logic present in api_get, but no recovery guidance in tool responses). No tool annotations (readOnlyHint, etc.). Security is adequate (no secrets in params, read-only operations), but no audit logging or permission gates.

• Document output schemas for all tools. Example: 'Returns array of {market_id, slug, title, yes_price, no_price, volume_24h, created_at}'. This enables LLMs to plan downstream calls and extract the right fields.
• Add min/max constraints to numeric parameters. Example: 'limit' should be 'minimum: 1, maximum: 100' in schema. 'period' should be 'minimum: 1, maximum: 365'.
• Add enum constraint to 'tag' parameter in list_events. Example: 'enum: ["crypto", "election", "sports", "economics", "science"]' instead of free-form string.
• Enhance tool descriptions with differentiation. Example: 'top_markets: Returns markets ranked by 24h trading volume (use for recent activity). top_movers: Returns markets with largest price changes (use for volatility analysis).'
• Add tool annotations to schema. Mark all tools with 'readOnlyHint: true' since they only read Polymarket data. This signals to LLMs that tools are safe to retry and call speculatively.

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

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