FNewsCrawler exposes two web-scraping tools via FastMCP with complete input schemas and descriptions. Both tools have clear, well-structured parameters with proper typing. However, there are notable gaps in error handling guidance, output schema documentation, and composition patterns. The tools follow a verb-noun naming convention but lack the granularity expected for production-grade agent tooling. Parameter descriptions are adequate (72 - 150 chars) but could be more LLM-optimized. Output schemas are not explicitly documented, forcing agents to infer structure. Error handling is minimal, no recovery guidance or error categorization. Security is reasonable (no credentials exposed as parameters), but logging/audit trails are not visible in the tool definitions. The server supports HTTP transport (via FastMCP framework) and includes cancellation support, which are positives for protocol readiness.

• Document output schemas for both tools. Include return type (object/list), field names, types, and what they represent. Example: 'Returns {success: bool, count: int, data: list[dict], raw_data: string, url: string, extracted_data: ...}'
• Add error handling guidance to descriptions. Example: 'Raises: SelectorNotFound (retry with verified CSS selector), Timeout (increase wait_timeout or check network), NetworkError (retryable)'
• Convert extract_type parameter to an enum: 'text' | 'html' | 'attribute' | 'mixed'. Add similar enums for pandas_* parameters if applicable.
• Document pandas parameter dependencies. Specify which combinations are valid and provide example usage: 'If using pandas_header, do not also use pandas_skiprows. If pandas_attrs is a JSON object, pandas_match is ignored.'
• Add result limit documentation. Example: 'Returns up to 500 matching elements. For pages with more, use CSS selector refinement or pagination via multiple requests with different offsets.'
• Clarify context_name semantics. Is it a browser session ID? Can agents create new contexts or must they use defaults? Add guidance: 'Use default context for single-page extractions. Use custom context names to maintain state across multiple pages on the same domain.'
• Rename context_name to session_name or browser_session_id for clarity and alignment with standard MCP parameter naming conventions.
• Add per-parameter validation constraints. Example: 'wait_timeout must be 100 - 60000 ms. css_selector must be a valid CSS 3 selector and should target at least one element.' Include these in parameter descriptions as well as schema constraints.

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

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