An MCP server that analyzes startup ideas through market research, competitor analysis, and social media trends. It integrates multiple MCP servers (market-data, serp-search, social-trends) to provide comprehensive startup viability assessments.
The server defines 2 tools with proper MCP registration and complete JSON schemas. Both tools have action-verb names and reasonable descriptions (156-167 chars). Parameter schemas are well-formed with types, ranges, and defaults. However, descriptions lack LLM-optimization guidance (e.g., when to use search vs search_news, what constitutes a good query). Output schemas are not documented, the tool descriptions do not explain the structure of returned JSON. Error handling returns JSON-formatted errors but lacks actionable recovery guidance for agents. The implementations show validation and logging, but the tool definitions themselves do not expose this rigor to the LLM.
Analyze social media trends and sentiment for a topic
Get financial information about public companies in a sector
Get market growth trends and projections
Get market size information for a specific industry or sector
Deep analysis of Reddit discussions on a topic
Search the web for information using Google Search API
Search for recent news articles on a specific topic
Output schemas not documented. Tool descriptions do not specify the structure of returned JSON (e.g., 'Returns a JSON object with "results" array containing objects with fields: position, title, link, snippet, displayed_link'). LLMs cannot plan downstream operations without knowing the response structure.
Error handling lacks actionable guidance. Errors are returned as JSON but do not guide the LLM on recovery steps. E.g., 'SERP_API_KEY environment variable is required' should be 'SERP_API_KEY not configured, ensure the server is initialized with a valid SerpAPI key from serpapi.com. Contact your administrator if you lack access.'
Tool descriptions lack disambiguation. Both tools perform searches but descriptions do not explain when to use 'search' vs 'search_news'. For example: 'Use this for current news and recent articles (hours to weeks old). Use search for general information, historical content, and archived articles.' This forces the LLM to guess.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 8 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 47 | - | v1 |
Analyze Twitter sentiment and engagement for a topic
Query parameter descriptions are minimal (10-17 chars). 'Search query string' and 'News search query' do not explain what constitutes a good query or how the tool interprets multi-word queries. Better: 'A search query (supports AND, OR, quoted phrases, site: filters). Example: "AI startup funding 2024" site:techcrunch.com'.
Result limiting not documented. Both tools cap results (num_results max 20, max 10) but descriptions do not explain why or what the LLM should expect. A note like 'Limited to 10 results per call to preserve token budget. Use pagination by making multiple calls with different page offsets if more results are needed.' would guide multi-call planning.