The server defines 3 tools with explicit schemas and descriptions visible in src/index.ts. All tools have non-empty descriptions (104-218 chars, within the 10-1024 baseline). Tool names follow verb_noun convention (search, search_advanced, login). However, there are significant gaps: (1) No output schema documentation, what fields do search results contain? What format are sources? (2) Parameter descriptions lack actionable constraints, search accepts any string with no guidance on format, length, or when to use search vs search_advanced. (3) search_advanced sources enum is documented but the multi-source behavior lacks clarity on how results are combined or prioritized. (4) No error handling guidance, what happens if Perplexity is unavailable, authentication fails mid-search, or timeouts occur? (5) login tool parameters are an empty object with no description of what triggers browser launch or what 'authenticated' means to the user. (6) The formatResult() function returns plain text; no structured output schema is exposed to the agent. (7) Missing dependency hints: search_advanced should state 'Use this when you need academic or social sources; prefer search for general queries', but doesn't guide when one is better. The server lands at the fair/poor boundary: descriptions exist but lack LLM optimization, schemas are incomplete (no output structure), and error recovery paths are absent.

• Document the output schema explicitly. Define: search/search_advanced return {answer: string, sources: [{title: string, url: string}], source_count: number}. login returns {authenticated: boolean, message: string}. Include in JSDoc or a separate schema document.
• Add LLM-optimized descriptions for all tools. Example for search: 'Search the web using Perplexity.ai and get an AI-synthesized answer with cited sources. Default Perplexity settings. Answer is always a single paragraph; sources are extracted from Perplexity's citations. Returns structured {answer, sources} JSON.' (Keep under 200 chars; be explicit about format.)
• Clarify search vs search_advanced. Describe search as 'Fast web-only search for general queries' and search_advanced as 'Combine web, academic, and social sources for research-heavy questions. Use when citations from papers or forum discussions are needed.' Add guidance: 'search() is 2-3x faster; use unless you specifically need academic sources.'
• Document parameter constraints for search: 'Query string (max 500 chars). Accepts natural language questions or keywords. Avoid very short queries (<3 chars) as Perplexity may return no results.' Add a note: 'Multi-word queries often work best, e.g., "latest advances in LLMs" vs "LLM".'
• Rewrite login description to be actionable: 'Verify authentication status and initiate login if needed. Checks if you have an active Perplexity.ai session; if not, opens a browser window for manual login. Waits up to 5 minutes for login completion. Returns success or timeout error.' Explain that authentication persists in the persistent browser profile.