Kagi MCP presents a single, well-engineered search tool with comprehensive parameter documentation and clear descriptions. The tool follows strong naming conventions ('kagi_search_fetch' is action-verb, domain-specific), has a detailed description that explains when to use it, and exposes an exceptionally rich parameter schema with proper typing, constraints, and per-parameter descriptions. All parameters include clear documentation of format, constraints, and usage context. Output structure is implicit in the implementation but not explicitly documented. No significant composition issues exist (single tool, read-only safety). Schema completeness is excellent, every parameter has type, constraints (min/max/enum), and descriptive text. Error handling is present in the code (proper exception handling, trace suffix formatting) but error guidance text is not explicitly documented in tool descriptions. No security issues detected, API key is handled server-side via Authorization header, not exposed as a parameter.

• Document the output schema in the tool description or in comments. Specify the structure of the returned results: e.g., 'Returns an object with a results array containing {title, url, snippet, date} fields, optional extracts array with full page content, and summary if available.'
• Add error guidance to the tool description. Example: 'If results are empty, the LLM should reformulate with broader keywords or fewer filters. If a timeout occurs, retry with a smaller limit. Invalid lens_id will return a clear error message.'
• Explicitly state parameter constraints in the tool description as a bulleted list or table. Example: '(lens_id is mutually exclusive with include_domains, exclude_domains, time_relative, file_type, use one approach or the other, not both.)'
• Consider adding a 'default_lens' or 'apply_default_lens' option to simplify common cases (e.g., academic search defaults to lens_id='2'). This would reduce parameter count for simple use cases while keeping advanced filtering available.
• Document timeout behavior and retry strategy. The code sets configurable KAGI_SEARCH_TIMEOUT and KAGI_EXTRACT_TIMEOUT; mention these in the description so clients understand retry expectations.