Single tool 'scrape' has a clear, concise description (89 chars) and well-defined input schema with 4 parameters, each typed and described. However, the tool lacks output schema documentation, critical for LLMs to plan downstream operations. No error handling guidance is provided; the function returns a dict with 'success', 'error', and 'status' fields, but the tool description does not document this structure or recovery paths. Parameter descriptions are adequate but lack format constraints (e.g., timeout range, format list examples). The tool name 'scrape' is generic and does not follow verb_noun convention (should be 'scrape_url' or 'fetch_url'). No tool annotations (readOnlyHint, idempotentHint) are present. Overall, the tool is functional but falls short of production-grade quality.

• Document the output schema in the tool description: 'Returns {success: bool, markdown: str, html?: str, error?: str, status?: int}. On success, markdown contains the cleaned content; on failure, error explains why (e.g., "Invalid URL", "Request timed out") and status is the HTTP code if applicable.'
• Rename tool from 'scrape' to 'scrape_url' to follow verb_noun convention and improve LLM intent recognition.
• Add error recovery guidance: 'If timeout occurs, retry with a longer timeout. If HTTP 429 (rate limit), wait before retrying. If HTTP 401, check that FASTCRAWL_API_KEY is valid.'
• Constrain 'formats' parameter: change description to 'Optional list of output formats. Valid values: ["markdown", "html"]. Defaults to ["markdown"].'
• Constrain 'timeout' parameter: change description to 'Timeout in seconds (1 - 300, default 30). If exceeded, returns {success: false, error: "request timed out after Xs"}.'
• Add tool annotations: mark as readOnlyHint=true (does not modify state) and idempotentHint=true (safe to retry).