MCP server providing zero-API-key web search infrastructure with real-time information retrieval, page browsing, claim verification, and evidence reporting capabilities.
Server has 7 well-named tools with complete input schemas and descriptions. Tool names follow verb_noun convention (search_web, browse_page, verify_claim, evidence_report). Descriptions are substantive (100-300 chars), explaining WHAT and WHEN. Parameters are typed with enums where appropriate (search type, region, profile, engine). However, output schemas are not documented, LLMs cannot predict response structure for downstream chaining. Error handling guidance is absent. Some parameter descriptions lack format/constraint details (e.g., 'query' has no length guidance). No tool annotations (readOnlyHint/destructiveHint) despite clear risk profiles.
Fetch and extract content from a URL. Returns Markdown by default (use format='text' for plain text). Supports caching, cross-host redirect detection, domain filtering, and PDF extraction. Use the prompt parameter to specify what information you want from the page.
Clear the response cache for fresh results.
Generate a citation-ready evidence report combining search, verification, and analysis.
List available search providers, provider profiles, goggles presets, and setup hints.
Build compact, citation-ready Markdown context for LLMs. Prefer this when an agent needs grounded context rather than raw search results.
Search the web for real-time information, news, images, books, or videos. Always use this to verify facts or get up-to-date information before answering. Optional providers include Bright Data when explicitly configured.
Output schemas not documented. LLMs cannot predict response structure (fields, types, pagination) for downstream tool chaining or data extraction.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). clear_cache is destructive but unmarked; search_web and browse_page are read-only but not annotated.
Error handling guidance absent. No recovery hints (e.g., 'If search fails, try a simpler query'). No error categorization (retryable vs. user-fixable vs. fatal).
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 69 | <=2025-11-25 | v2 |
Evaluate whether a factual claim is supported, contested, or under-evidenced.
Parameter descriptions lack format/constraint details. 'query' has no length guidance; 'max_chars' has no min/max bounds; 'laya_threshold' has no range (0.0 - 1.0 assumed but not stated).
clear_cache description is minimal (11 words). Does not explain side effects, when to call it, or what 'cache' means (response cache? provider cache? both?).