The server defines 2 tools with clear verb-noun naming (firecrawl_scrape, firecrawl_map). Both tools have descriptions and input schemas with parameter details. However, there are significant gaps: (1) No documented output schemas, callers cannot determine what fields to expect; (2) Parameter descriptions lack constraint details (valid ranges, format specifications, regex patterns); (3) No error handling guidance in descriptions or tool docs; (4) Missing documentation on when/why to use each tool vs alternatives; (5) No mention of pagination limits or result size caps despite tools potentially returning large datasets. The tool factory and error handler show good engineering patterns, but tool definitions themselves need deeper LLM-optimization work. Average of individual tool scores: (firecrawl_scrape: 63 + firecrawl_map: 61) / 2 = 62.

• Document the output schema for both tools. For firecrawl_scrape, specify: 'Returns { content: string, metadata: { title, description, author, ... }, format: string }'. For firecrawl_map, specify: 'Returns { urls: [ { url, title?, depth?, ... } ], totalCount, hasMore }.'
• Add explicit min/max constraints to numeric parameters. E.g., 'timeout: 1000 - 120000 milliseconds (default 30000)' and 'waitFor: 0 - 60000 milliseconds (0 = no wait)'.
• Document each formats enum value. E.g., 'markdown: Converts HTML to Markdown (recommended for LLM processing). rawHtml: Returns unmodified HTML (useful for custom parsing). screenshot: Returns a Base64-encoded PNG screenshot (useful for visual content).'
• Add error handling guidance to tool descriptions. E.g., 'If the URL is unreachable (connection timeout, 404, 403), the tool returns an error. Retry with exponential backoff or suggest the user verify the URL. If JavaScript rendering times out, try increasing waitFor or reducing page complexity.'
• Explain when to use each tool in their descriptions. E.g., firecrawl_scrape: 'Use this to extract structured content (text, links, metadata) from a single webpage. Returns content in your choice of formats.' firecrawl_map: 'Use this to discover all pages within a website's structure. Returns a list of URLs without detailed content. Call this first to determine scope; then use scrape to extract details from specific pages.'
• Implement pagination or result limiting with guidance. E.g., add a 'maxResults' parameter capped at 100 to firecrawl_scrape (for multiple formats). For firecrawl_map, document limit=5000 as the maximum and recommend starting with limit=100 for safety, then increasing if needed.

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

• Relies on Sampling (deprecated) - integrate directly with the LLM provider API