Loading ToolBench
Official Model Context Protocol server for Jina AI APIs. Provides web search, URL reading, screenshot capture, academic paper search, image search, and semantic reranking capabilities.
Server has 19 tools with consistent Zod-based schemas and descriptions. Naming follows verb_noun convention (search_web, read_url, capture_screenshot_url). Descriptions are present and contextual (avg ~150 chars), meeting the 10-1024 char baseline. However, critical gaps exist: (1) parameter descriptions are minimal or absent for many optional params (e.g., 'tbs', 'gl', 'hl' in search_web lack guidance on format/values); (2) output schemas are undocumented, no structured response format specified for any tool, forcing LLMs to infer result structure; (3) error handling is generic ('Error: ...') with no recovery guidance or categorization; (4) no pagination guidance despite tools returning 'num' results up to 30; (5) tool show_api_key exposes bearer tokens in responses, violating secret-injection pattern. Naming is strong, descriptions adequate, but schema documentation and error handling are weak.
Capture high-quality screenshots of web pages in base64 encoded JPEG format. Use this tool when you need to visually inspect a website, take a snapshot for analysis, or show users what a webpage looks like.
Remove duplicate and near-duplicate images from a list using visual similarity. Returns unique images with diversity optimization.
Remove duplicate and near-duplicate text items from a list using semantic similarity. Returns unique items with diversity optimization.
Expand a search query with semantically related terms and synonyms to improve search coverage and relevance. Returns the original query plus expanded terms.
Extract structured content from PDF files including text, tables, figures, and equations. Returns organized content with metadata.
Output schemas undocumented. No tool specifies what fields are returned, forcing LLMs to infer result structure. This violates pattern:tool and pattern:response-shaper, increasing hallucination risk and breaking downstream tool chaining.
Optional parameter descriptions missing or vague. Parameters like 'tbs' (time-based search), 'gl' (country code), 'hl' (language) lack format guidance, valid values, or examples. LLMs cannot infer correct usage.
show_api_key tool returns bearer tokens in response content, violating pattern:secret-injection. Tokens should never appear in tool responses, they enter LLM context and risk exposure in logs or user-facing output.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | 2026-07-28+ | v2 |
Guess the last updated or published datetime of a web page. This tool examines HTTP headers, HTML metadata, Schema.org data, visible dates, JavaScript timestamps, HTML comments, Git information, RSS/Atom feeds, sitemaps, and international date formats to provide the most accurate update time with confidence scores. Returns the best guess timestamp and confidence level.
Read and extract content from multiple URLs in parallel for improved performance. Returns markdown content for all URLs.
Execute multiple arXiv searches in parallel for improved performance. Returns results for all queries.
Execute multiple SSRN searches in parallel for improved performance. Returns results for all queries.
Execute multiple web searches in parallel for improved performance. Returns results for all queries.
Get up-to-date contextual information of the current session to provide localized, time-aware responses. Use this when you need to know the current time, user's location, or network environment to give more relevant and personalized information.
Extract and convert web page content to clean, readable markdown format. Perfect for reading articles, documentation, blog posts, or any web content. Use this when you need to analyze text content from websites, bypass paywalls, or get structured data.
Search arXiv for academic papers and research articles. Returns relevant papers with titles, abstracts, and URLs.
Search for images online. Returns image results with URLs and metadata.
Search the Jina AI blog for articles and news. Returns blog posts with titles, excerpts, publication dates, and URLs.
Search SSRN (Social Science Research Network) for academic papers in economics, finance, law, and social sciences. Returns relevant papers with titles, abstracts, and URLs.
Search the web for information using Jina AI's web search API. Returns relevant search results with titles, snippets, and URLs.
Return the bearer token from the Authorization header of the MCP settings, which is used to debug.
Rerank and sort search results or text items by semantic relevance to a query. Uses embedding-based similarity scoring.
Error handling is generic ('Error: ...') with no recovery guidance. Errors do not categorize as retryable, user-fixable, or fatal, and do not suggest next steps. Violates pattern:recovery-guide.
Pagination not documented. Tools accept 'num' parameter (default 30) but do not specify max limits, whether results are paginated, or how to fetch next page. Violates pattern:paginated-result.