Server provides 6 tools with consistent naming (verb_noun pattern: web_search, web_fetch, research_*) and reasonable descriptions (80-150 chars). All tools are READ_ONLY, reducing security risk. However, input schemas are inferred from test files rather than directly visible in source code, and parameter descriptions lack detail on constraints, formats, and valid ranges. No output schemas documented. Error handling guidance absent.

• Expose tool definitions directly in server.py or a dedicated tools module with explicit @server.list_tools() registration. Include full JSON Schema for all parameters with type, description, enum, minLength, maxLength, minimum, maximum, pattern, and default where applicable.
• Expand parameter descriptions to 50-100 chars each. For 'mode' in web_search, add: 'Search mode: auto (balanced), fast (speed), instant (cached), deep-lite (thorough), deep (very thorough), deep-reasoning (reasoning-optimized). Default: auto.'
• Document output schemas for each tool. E.g., web_search returns {results: [{title, url, snippet, source}], total_count, next_cursor}. web_fetch returns {content, url, extracted_text, metadata}.
• Add error handling guidance to descriptions. E.g., 'Returns 429 if rate-limited (retry after 60s). Returns 404 if URL unreachable (try web_search to find alternative). Returns 400 if query malformed (check syntax).'
• Add bounds and constraints to numeric/enum parameters. E.g., max_chars: 'Maximum characters to return (1 - 100000, default 10000). Larger values increase latency and cost.'

Inferred effective spec: 2025-06-18+.

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