Server provides 8 tools with consistent naming (verb_noun pattern: hn_top, hn_search, hn_discussion, arxiv, lobsters, github_trending_ai, tech_signal_digest, hn_best). All tools have descriptions (50-100 chars, within baseline 34-392 range). Input schemas are present with typed parameters and descriptions. However, output schemas are not documented, LLMs cannot predict response structure for planning downstream calls. Parameter descriptions are minimal (10-20 chars) and lack format/constraint details. No error handling guidance visible. Tools follow single-responsibility principle well (each targets one data source or aggregates them). Missing: pagination guidance, rate limit documentation, and response field documentation.

• Document output schema for each tool. Example for hn_top: 'Returns array of {id: int, title: string, url: string, score: int, by: string, time: int, descendants: int}'. This enables LLMs to extract fields and chain calls.
• Expand parameter descriptions to include constraints. E.g., 'limit: Maximum number of stories (1-100, default 30). Higher values increase latency; recommended ≤50 for real-time chat.'
• Add error handling guidance to tool descriptions. E.g., 'If no results found, try: (1) broaden query terms, (2) increase limit, (3) check if source is available. Retryable on 5xx/429.'
• Declare tool annotations (readOnlyHint=true for all tools) in MCPServer registration to signal safety and enable agent optimizations.
• Document pagination strategy. E.g., 'Results capped at limit parameter. For more results, use offset-based pagination: call again with offset=limit to fetch next page.'