Server has 9 tools with basic schemas and descriptions, but significant quality gaps. Tool naming is inconsistent (perplexity_ask, perplexity_ask_v2, perplexity_search, search, redundant variants). Descriptions are generic and under 50 chars for most tools, lacking WHEN/WHY context. Parameter descriptions exist but are minimal (e.g., 'The question to ask' for query). No output schemas documented. No error handling guidance. No tool annotations (readOnlyHint, etc.). Composition is poor: 3 search variants + 2 research variants + 2 ask variants suggest the agent must guess which to call. Files parameter is typed as 'object' with no schema detail. No pagination, limits, or result structure guidance.

• Consolidate redundant tools: merge perplexity_ask + perplexity_ask_v2 into a single 'ask_perplexity' tool with a 'stream' boolean parameter. Same for research and search variants. Reduces cognitive load on LLM.
• Expand tool descriptions to 80-150 chars. Example: 'Ask Perplexity a question in auto mode. Use this for quick answers with web search. For deep analysis, use ask_perplexity_reasoning instead. Returns answer text and source citations.'
• Document output schemas for each tool. Example: 'Returns {answer: string, sources: [{title, url, snippet}], model_used: string}'.
• Replace 'files' parameter object with explicit fields: 'file_paths: [string]' with description 'Paths to local files to upload (max 5, max 10MB each)' and 'file_types: [enum: pdf|txt|docx|csv]'.
• Add error handling guidance to descriptions: 'On rate limit, wait 60s before retry. On invalid query, returns 400 with suggestion. On network error, returns 503, retryable.'

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

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