Two tools with basic schemas and descriptions, but significant gaps in parameter documentation and output structure. firecrawl_scrape has a well-defined input schema with enum constraints for formats, but lacks output schema documentation. firecrawl_search has minimal parameter description and no documented output structure. Neither tool includes error recovery guidance or actionable error messages. Naming is verb-noun compliant but descriptions are generic (194 chars baseline; these average ~120 chars). No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite both being read-only operations.

• Add formal output schema documentation for both tools. Example for firecrawl_scrape: 'Returns { content: [{ type: "text", text: string }] }. For markdown format, text contains full page markdown. For links format, text contains newline-separated URLs.'
• Expand firecrawl_search query parameter description to: 'Search query string (1-500 chars). Supports natural language and boolean operators. Returns up to 20 results ranked by relevance.'
• Add error recovery guidance: 'If scrape fails with timeout, retry with a simpler format (markdown only). If API key is missing, configure FIRECRAWL_API_KEY environment variable.'
• Add readOnlyHint annotations to both tools via server.tool() metadata to signal they are safe to retry and do not modify state.
• Document firecrawl_scrape formats default: 'If omitted, defaults to ["markdown"]. Combine multiple formats to get different content representations in a single call.'