Loading ToolBench
A proxy MCP server that bridges Tavily Search API and Brave Search API with support for multiple deployment options (HTTP bridge server, stdio bridge, and Cloudflare Workers)
MCP Nexus exposes 7 search and content extraction tools with good parameter coverage but inconsistent quality across the toolkit. All tools have descriptions and properly typed schemas with enums and constraints. However, output schemas are entirely undocumented, LLMs cannot know what fields to expect from search results or extraction responses. Naming is clear and action-oriented (tavily_search, brave_web_search). Parameter descriptions are moderately detailed but lack guidance on when to choose between competing tools (e.g., tavily_search vs brave_web_search vs tavily_research). Error handling is minimal, no guidance on retryability, rate limits, or recovery paths. The tools are READ_ONLY (safe from destructive operations) but lack error categorization and actionable failure messages. Overall, definitions are better than median community servers but fall short of production-grade quality due to missing output documentation.
Search for local businesses and places using the Brave Search API. Implementations commonly fall back to web search if local results are unavailable. Returns a JSON array of results.
Performs a web search using the Brave Search API. Use for general web searches for information, facts, and current topics. Returns a JSON array of results.
Crawl a website starting from a URL. Extracts content from pages with configurable depth and breadth.
Extract content from URLs. Returns raw page content in markdown or text format.
Map a website's structure. Returns a list of URLs found starting from the base URL.
Perform comprehensive research on a given topic or question. Use this tool when you need to gather information from multiple sources to answer a question or complete a task. Returns a detailed response based on the research findings.
No documented output schemas for any tool. LLMs cannot infer what fields search results contain (e.g., does tavily_search return 'url' or 'link'? 'title' or 'headline'? 'snippet' or 'summary'?). This forces agents to guess at field names and risks malformed downstream calls.
No guidance on tool selection. tavily_search, tavily_research, and brave_web_search all perform web search but with different tradeoffs (depth, model complexity, speed). Descriptions do not explain when to prefer one over another, forcing LLMs to reason by trial or pick arbitrarily.
Missing error handling guidance. No indication of what errors are retryable (rate limit, timeout) vs user-fixable (invalid query) vs fatal (API key missing). Agents cannot distinguish between transient failures and permanent problems.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 62 | 2026-07-28+ | v2 |
| 2026-03-09 | D | 59 | - | v1 |
Search the web for current information on any topic. Use for news, facts, or data beyond your knowledge cutoff. Returns snippets and source URLs.
Parameter descriptions lack format constraints and examples. For example, 'start_date' says 'YYYY-MM-DD' but does not clarify if it must be in the past, whether relative dates (e.g., '7 days ago') are accepted, or what happens if start_date > end_date. Underdocumented constraints invite LLM input errors.
brave_web_search and brave_local_search have 'additionalProperties: true' in their schemas, which signals undefined extra parameters. This may indicate incomplete schema definitions or overly permissive parameter acceptance, risking silent parameter mishandling.
tavily_extract, tavily_crawl, and tavily_map descriptions do not specify pagination or result limits. Agents do not know if these tools cap results at N items or whether they support pagination for large crawls. Without limits documented, agents cannot safely compose multi-page workflows.